Null-Subject Properties of Slavic Languages

Home , Bohemians (tribe), Czech language, Domowina, Lower Sorbian language, Sorbian languages, Starosta

Slavistische Beiträge ∙ Band 361

(eBook - Digi20-Retro)

Martina Lindseth

With Special Reference to Russian, Czech and Sorbian

Verlag Otto Sagner München ∙ Berlin ∙ Washington D.C.

Digitalisiert im Rahmen der Kooperation mit dem DFG-Projekt „Digi20“ der Bayerischen Staatsbibliothek, München. OCR-Bearbeitung und Erstellung des eBooks durch den Verlag Otto Sagner: http://verlag.kubon-sagner.de

«Verlag Otto Sagner» ist ein Imprint der Kubon & Sagner GmbHMartina. Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

S lavistische Beiträge

Begründet von Alois Schmaus

Herausgegeben von Peter Rehder

Beirat: Tilman Berger • Walter Breu • Johanna Renate Döring-Smimov Wilfried Fiedler • Walter Koschmal • Ulrich Schweier • MiloS Sedmidubskÿ * Klaus Steinke

BAND 361

V erla g O tto Sa gner Martina Lindseth - 9783954790661 M ünchen Downloaded 1998 from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

Martina Lindseth Null-Subject Properties of Slavic Languages With Special Reference to Russian, Czech and Sorbian

V erla g O tto S a g n er Martina Lindseth - 9783954790661 M ünchen 1998Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access PVA 98. 2677

ISBN 3-87690-699-7 © Verlag Otto Sagner, München 1998 Abteilung der Firma Kubon & Sagner D-80328 München

Gedruckt auf alterungsbeständigem PapierMartina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

ACKNOWLEDGEMENTS

This work originates from my Ph.D. thesis which I defended at Indiana University, Bloomington, U.S.A. in 1996. I would like to thank all members of my dissertation committee, especially Dr. Steven Franks, for guiding me through my graduate studies at IU and the preparation of my dissertation. I would also like to express my graditute to the scholars at the Serbski institut in Budyšin. Special thanks go to Dr. Irene Šērak for helping me find native speakers for my fieldwork, and for reading portions of the manu-

script, and to Dr. Helmut Faßke for allowing me access to some of his recent data. I would also like to thank Erik for all his support and patience while I was working on the dissertation, and Pat for proofing the book manuscript. This research was supported in part by an IU International Programs Pre-Dissertation Fellowship, an IU College of Arts and Sciences Summer Research Fellowship, a Wabash College Faculty Devel- opment Grant and an American Council of Learned Societies Disserta-

tion Fellowship. All financial help was greatly appreciated and allowed me to expand the project far beyond my original goals. This book is dedicated to my father. Dr. Siegfried Profe, who always rejoiced in my successes. He would have celebrated the com- pletion of this work.

Eau Claire, Wisconsin, summer 1998

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access TABLE OF CONTENTS ACKNOWLEDGEMENTS

1. INTRODUCTION...... 1 I .X. The issue of null nubjects 1.2. Slavic languages and universal grammar 1.3. Organization

2. NULL SUBJECTS IN THE GOVERNMENT BINDING FRAMEWORK...... 8 2.1. Theoretical background 2.2. Are null subjects more than just nothing? 2.3. Development of the pro-drop parameter 2.4. Null-subject diagnostics 3. NULL - SUBJECT CHARACTERISTICS OF SLAVIC LANGUAGES...... 4 0 3.1. Application of null-subject diagnostics 3.1.1. West Slavic and South Slavic 3.1.2. Russian (East Slavic) 3.2. Why is Russian different? 3.2.1. The intuitive explanation 3.2.2. How to formalize the difference between Russian and West and South Slavic 3.3. Summary

4. IS SORB IAN A Pro-DROP LANGUAGE? ...... 73 4.1. The Sorbian population of Germany 4.2. Verbal morphology of Upper Sorbian 4.3. Grammar vs. Reality 4.4. Application of null-subject diagnostics 4.5. Pronouns and purism 4.6. Why is Sorbian not a canonical null-subject language? 4.6.1. Identification and the number of syncretisms 4.6.2. A word order account 4.6.3. Challenges for the identification/licensing framework 4.6.4. The competition model and the loss of pro-drop

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 000Б2018

-VII- 4.7. Lower Sorbian 4.8. Summary 5. THE STATUS OF OVERT EXPLETIVES IN CZECH AND SORBIAN... 14 4 5.1. Licensing/identification and the typology of languages 5.2. Spoken Czech 5.2.1. Syntactic status of overt expletives 5.2.2. Origin and meaning of Czech expletives 5.3. Upper Sorbian 5.3.1. Syntactic status of overt expletives 5.3.2. Historical perspective 5.3.3. Upper Sorbian and German 5.4. Summary 6. CONCLUSION ...... 196 BIBLIOGRAPHY ...... 198

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access ־׳ fA

■דד te ^ I& ísz 'VUÍV^* Ч-5Й

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־1־ 1. INTRODUCTION 1*1• The issue of null subjects

It is readily apparent that languages vary according to whether an unstressed pronominal subject of a finite clause is lexicalized (English, French) or not (Spanish, Italian). This

study derives from and contributes to the large corpus of recent research by generative grammarians which has sought to explain this crosslinguistic variation in the occurrence of null subjects. In a broader sense this work is part of an even larger tradition since generative syntacticians were certainly not the first to concern themselves with the zero-problem in linguistics. Indeed, the concept of zero-elements (also called null elements or phonologically empty elements) has been previously invoked by many linguists belonging to different schools of linguistics, including those of the French linguist de Saussure and his students, Jakobson and the Prague functio- nalists, and Bloomfield and the American Structuralists. As a result Meier (1961) was able to devote an entire monograph in which he included ,״to the "Zéro-Problem in der Linguistik a survey of "Zéro-Spezial-Arbeiten" and a survey of the histo-

ry of this topic in linguistics. For generative linguists, null subjects have proven to be a fruitful topic of debate ever since Perlmutter's Deep and surface structure constraints in syntax (1971). Since one of the central concerns of the "principles and parameters" model

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-2- of government and binding (GB) theory has been to determine

positions in which overt and phonologically empty manifesta-

tions of different NP types can occur, defining the precise

nature of the crosslinguistic variation in the availability of

empty subjects has been of special interest to GB, generating numerous articles and several books, such as Bouchard (1984), Roberge (1990), and Jaeggli and Safir (1989). Traditionally, Italian has been cited as an example of

the canonical null-subject or pro-drop languages which are clearly identifiable as different from canonical overt subject languages, such as English, as illustrated in (1).

(1) parlo I speak parli you speak parla he speaks

parliamo we speak parlate you speak parlano they speak

Since the Italian inflectional paradigm uniquely distinguishes all person/number combinations, it is intuitively clear, independent from any specific grammatical framework, why Italian

can afford to do without overt subject pronouns: The pronominal features (person, number) of a "missing" subject can be recovered unambiguously due to the rich inflectional agreement on the verb.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access However, with the eventual spread of theoretically in- formed generative studies to more and more languages, it became apparent that the situation is far more complex than expected. It is now clear that the relationship between rich agreement and null subjects is less direct than originally assumed. On the one hand, there are languages like German, which seem to have what we would call rich agreement, but fail to allow referential null subjects, while on the other hand, there are languages like Japanese and Chinese, which are quite capable of omitting subject pronouns, despite their lack of agreement morphology. Further research revealed that some languages (such as German, Icelandic and Russian) can show mixed behavior with respect to the pro-drop parameter. In German, for instance, referential subjects and quasi-argumental1 subjects are not omitted, whereas truly expletive subjects are.

(2) ich/*0 liebe Tiere '(I) love animals1 (3) draußen regnet es/*0 schon wieder ,outside it is raining again* (4) ich bin froh, daß 0/*вв getanzt wird 1I am happy that there is dancing going on'

1 As pointed out by Chomsky (1981: 325), quasi-arguments have some properties of true arguments (e.g. control of PRO) and some properties of non-arguments (cannot be replaced by referential NPs) and thus behave differently than truly expletive subjects.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-4- Thus there is in fact no clearcut division between null- subject and non-null-subject languages. Despite the longstanding intuition that there is a relationship between richness of agreement and the availability of null subjects in a specific grammar, a definitive and completely accurate characterization of the notion of inflectional richness has yet to be proposed.

1.2. Slavic languages and universal grammar

The theoretical parts of this work are conceived within the syntactic framework of GB theory, initiated by Chomsky (1981,

1982) as a development of earlier versions of generative grammar and further refined during the past decade. The crucial hypothesis adopted by Chomskian syntacticians is that a large part of a speaker's knowledge about language is innate. All

humans are endowed with the same universal grammar (UG), which consists of a set of absolute universais and principles, which do not vary from one language to the next, and a range of parameterized principles whose actual values have to be fixed during childhood language acquisition upon exposure to lan- диаде. Thus, while the language-specific properties themselves are not fully determined by UG, the available options are. The grammar of a specific language boils down to a particular set of values for the parameters. This theory, for apparent reasons, is often referred to in the relevant literature as the "principles and parameters" theory and we shall use this term synonymously with "government and binding".

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Until fairly recently the Slavic languages received relatively little attention within GB theory. This is actually quite surprising, considering that these 14 closely related languages2 show interesting syntactic differences which can provide Chomskian syntacticians with an ideal pool of data to test their theories of universal principles and potential parametric variation. Certainly, from the perspective of null- subject phenomena, which are the focus of this work, the Slav- ic languages are quite diverse. The identification of particular problems in the Slavic

languages may thus indicate modifications which must be made

to current theories in general. Since it is a basic tenet of UG that any analysis proposed for some language(s) should be able to explain data of other languages and accommodate cross- linguistic variation, it follows that the investigation of data from "overlooked" languages is often crucial in proving or disproving the general validity of theoretical assumptions. In this context perhaps the most neglected of the modern Slavic languages is Sorbian, also referred to as Lusatian and Wendish, which belongs to the West Slavic language family. Since it is one of the most complex Slavic languages with regard to null subjects, it is a major focus of this present

2 East Slavic: Russian, Ukrainian, Belorussian West Slavic: Polish, Kashubian, Upper Sorbian, Lower Sorbian, Czech, Slovak South Slavic: Serbian, Croatian, Slovene, Macedonian, Bulgarian There is some disagreement among linguists whether Kashubian and Polish, and Serbian and Croatian are separate languages.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access work. Sorbian is generally accepted to have split into two different literary languages, identified as Upper and Lower Sorbian, which are spoken by a very small minority in and around the cities of Bautzen and Cottbus in Eastern Germany. The total assimilation of this Slavic society into the vastly greater German one which surrounds it also plays a role in this study, culturally as well as linguistically.

The rapid decrease in the number of native speakers of Sorbian in recent decades (about 22,500 remain) has been ac* celerated by the economic and social disruption brought about by the collapse of East Germany in 1989. It is hoped that this current research will help to slow this decline by in- creasing awareness, and contributing to the (re)introduction of the Sorbian languages into academic discussion. Ironical- ly, the same political changes which threaten the survival of the Sorbian language and culture have also improved access to native speakers of Sorbian by Western linguists.

1.3. Organization This work is divided into four main chapters, the first of which. Chapter 2, sets out to establish a theoretical background. The licensing and identification approach to null subjects is summarized and a set of three diagnostics introduced which can be used to decide whether or not a language has all the properties of a canonical null-subject language.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 000Б2018

-ד-

In Chapter 3, these diagnostics are applied to Slavic in general, identifying a split between East Slavic on the one hand, and West and South Slavic on the other. This is followed by a section which discusses why the Slavic languages are divided along these lines. Chapters 4 and 5 are devoted to Upper Sorbian and Czech. Both chapters present and discuss linguistic data which seem to challenge the licensing/identification framework. This begins with Upper Sorbian which provides a telling counter-

example to the long accepted intuition that the explicitness of verbal morphology and occurrence of null- pronominal subjects are directly related. A broad examination of contemporary and early texts reveals a surprisingly high level of overt pronoun use, which makes Sorbian an exception among the West Slavic languages. Chapter 5 focuses on the comparison and analysis of the occurrence of overt expletive subjects in Czech and Upper Sorbian. Czech generally omits unstressed pronominal subjects. This is expected, since Czech verb forms uniquely distinguish the different person/number combinations, very much like in the Italian example above. However, overt non-referential pronouns occur frequently in spoken Czech, which has generally been assumed impossible for languages which

omit referential subjects. The analysis of Czech expletive subjects is followed by a discussion of overt non-referential pronouns in Upper Sorbian.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 2. NULL SUBJECTS IN THE GOVERNMENT BINDING FRAMEWORK 2.1• Theoretical background

GB theory describes knowledge of language as a highly structured theory of UG which is composed of an interrelated set of six modular subtheories. These are enumerated by Chomsky (1981: 5) as bounding theory, government theory, theta-theory, binding theory, Case theory and control theory. These sub- systems, in turn, are made up of general principles and parameters whose potential value settings vary within clearly defined limits from one language to another. GB is also modular in the sense that it assumes four different levels of representation, known as D(eep)-structure, S(urface)-structure, phonetic form (PF) and logical form (LF). These different levels of representation are related to each other by a transformational component which contains the rule Move-a, which, in effect, allows anything to move anywhere. Various general principles regulate the mapping from one level to another by determining which movements are well-formed and which are not.1 One important constraint on the relation between syntactic representations is the structure-preserving

1 There are central structural concepts which are not confined to any particular subtheory, most notably the no- tions of "c-command" and "government". Here, we assume the following definitions (simplified from Chomsky, 1986): C-command (extended to include m-command): Node A c-commands node В iff A does not dominate В and every X that dominates A also dominates B. Government : A governs В iff A c-commands В and no barrier intervenes between A and B.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access ־9־ principle, derived from the work of Emonds (1976), according to which movement operations must not destroy existing structures. This means that those structures established at D- structure must be preserved at S-structure. It follows that when a given element is moved from its extraction-site to its landing-site it cannot simply disappear from its original site "without a trace". Hence, whenever an element is moved, a coindexed, empty element called a trace remains in the moved element's original position.

Theory-internal reasons force this syntactic framework to assume the existence of non-overt, but structurally present categories. Two principles, the theta-criterion and the projection principle, which define how the properties of lexical entries are projected onto syntax, are especially crucial in this respect. The theta-criterion specifies that each argument bears one and only one theta-role and that each theta- role is assigned to one and only one argument. According to the projection principle, the specifications for each lexical item given in its entry in the lexicon have to be met at all levels of syntactic representation. In other words, the one- to-one pairing between the thematic roles and arguments of a verb's predicate-argument structure required by the theta- criterion has to remain consistent on all levels of derivation (D-structure, S-structure, and LF) . The projection principle and the theta-criterion determine whether an NP has to be present in a syntactic representation. However, a structurally

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access required NP can be manifested either as phonologically realized (overt) or поп-realized (null). The latter is called an empty category. The various sub-theories of GB interact to determine whether the NP is lexical or not, and what type of NP appears in any given position. Certain empty categories can only appear in certain structural positions, which is reflected in the formulation of licensing requirements for empty categories. However, the exact formulation of such licensing conditions is far from being uniformly agreed upon. Furthermore, the content of those "missing" elements has to be

identifiable. The position of traces, for instance, is formally licensed by the empty category principle,2 while their content is determined by their antecedent. An interesting issue in this context is whether formal licensing of the occurrence and the identification of the content of an empty category are achieved by one and the same or two distinct operations. We shall return to this question in section 2.3. The typology of empty categories as developed by Chomsky (1982: 78-89) mirrors that of overt categories. The module of the grammar regulating the interpretation of NPs is the bind-

2 Traces must be governed in a special way in order to occur in certain positions: They must be properly governed. Empty category principle: Traces must be properly governed. A properly governs В iff A theta-govems В or A antecedent-governs B. (Chomsky, 1986: 17) . A head A theta-governs В if it governs and theta-marks B. A maximal projection A antecedent-governs В if A and В are coindexed and A governs B.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access ing theory. Chomsky proposes that the three overt NP-types, anaphor, pronoun and R-expression, should be interpreted as

נ.[composed of two features: [!anaphoric] and [!pronominal Depending on its feature matrices, the NP is subject to different principles of the binding theory. An NP-type which is subject to principle A of the binding theory has the feature [+anaphoric] and consequently has to be bound in its governing category. A category which is subject to principle В of the binding theory has the feature [+pronominal] and therefore has to be free in its governing category. Chomsky further assumes

that the typology of empty categories can also be derived from

the possible combinations of [*anaphoric] and [!pronominal]. The inventory of overt and empty NPs is illustrated in

Table 1

Table 1: Classification of NPs

Anaphoric Pronominal NPs Empty NPsOvert anaphors+ NP-traces + pronouns pro + PROexcluded R-expression Wh-traces (variables)

-This is analogous to the representation of syntac נ tic categories such as N, V, P and A by feature matrices rather than primitives.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access According to Table 1, anaphors and NP-traces are subject to principle A of the binding theory and must be bound within their governing category. Pronouns and pro have to be free within their governing category, and referring expressions and variables are subject to principle C, that is they must not be bound anywhere. The features proposed for the NP-types are binary, and, in the ideal case, we expect all four combinations to occur. The inventory of empty categories supports this expectation. However, there are only three overt NP- types. The occurrence of an overt NP-type with the feature composition [+anaphoric,+pronominal] is ruled out by independent principles of UG.4 The different overt and null elements are exemplified in the following sentences. The Slavic examples are from Russian unless otherwise indicated:

An element which is [+anaphor] must be bound within its governing category, in keeping with principle A of the binding theory. An element which is [+pronominal] must be free in its governing category in keeping with principle В of -anaphoric, +pro־♦-] the binding theory. NPs with the features nominal] are thus subject to contradictory requirements. The only way for an element to fulfill both principle A and В at the same time is to lack a governing category. If there is no governing category, then neither principle A nor principle В apply, because the domain in which the element has to be both free and bound is not defined. This can be fulfilled under the condition that the NP lacks a governor. However, an overt NP that lacks a governor cannot be case-marked, under the assumption that case marking requires government, and hence will be ruled out by virtue of the case filter, according to which all overt NPs have to be case marked. As an empty category, PRO is not subject to the case filter and the absence of a governor is not problematic. Furthermore, due to its feature composition PRO is only admitted in ungoverned positions. This proposition is referred to in the literature as the "PRO theorem".

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access (1 ) [♦anaphoric,-pronominal] Ove rt : anaphor Ivan, ljubit tol 1ko sebjaį 1 John* loves only himself/

Empty: NP-trace dom* byl postroen t, ,the houseį was built t/

(2 ) t-anaphoric,+pronominal] Overt : pronoun Ivan, nenavidet egoy.į 1 John, hates hia^•/ Empty: pro

pro miluju Mariu [CZ] 1 (I) love Mary* (3) [♦anaphoric,+pronominal] Overt: excluded Empty: PRO on xoCet PRO ženit'sja na Marii 'he wants [PRO to marry Mary]' (4) [-anaphoric,-pronominal] Overt: R-expression

0 П1 durnai, čto Ivan^ pobedil v zabege 'hej thought that John,/•* won the race1 Empty: variable kogoj vy priglasili tt ,whom, did you invite t/

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access A number of issues connected with the types of empty categories described above remain controversial, although the study of empty categories with respect to their type, distribution and contents has become a very popular topic within GB.

2.2• Are null subjects more than just nothing? Let us review some of the empirical and theory-internal reasons which support the assumption that empty subjects of both tensed and infinitival clauses are real in the sense that they are structurally present in a representation. Following

Chomsky (1986) I assume that X-bar theory, including the endo- centricity requirement for phrases, applies to both lexical and non-lexical (functional) categories. The clausal category traditionally labeled S is reinterpreted as the maximal pro- -INFL, and the category S' is reinter־jection of I, where I preted as the maximal projection of C, where C*complementizer.

(5) S ‘ (6) CP Comp SpecS IPСNP INFL VP IPСNP

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Unless otherwise stated, I will assume that structural subjects are base-generated in and occupy the specifier position of IP [NP, IP] .5

In the following I argue against (7a) and in favor of structure (7b) for all sentences, even if the subject is not phonologically realized:

(7) a. b. IP IP y 1 'v 1 1 NP 1 1 I VP I VP

In the sentences below it is obvious that even in the absence of an overt subject its presence is felt implicitly and its content can (and must) be recovered to ensure a complete statement. (Here and in later examples I shall use в to indicate the presence of a generic empty category)

(8) e biju ho [Cz] '(I) beat him' (9) on xočet [e bit' Ivana] [Ru] 1he wants to hit Ivan1

The verb ,to hit' takes two arguments: an external argument which receives the theta-role of agent, and an internal argu-

5 An alternative proposal which has been gaining ground in the recent literature is the so-called VP-internal subject hypothesis. Sportiche (1988) proposes that all subjects are base-generated in [Spec.VP] position, so that NPs in [Spec,IP] are derived subjects. I shall return to this hypothesis in subsequent chapters.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

16-־ ment which is assigned the role of patient. It follows from the theta-criterion in conjunction with the projection principle that a sentence with a verb which assigns an external theta-role must have a structural subject at any level of the representation-

Within GB, the missing subjects in (8) and (9) represent two fundamentally different types of empty categories, although they have in common the fact that they are base-generated, as opposed to the [-pronominal] empty categories, which represent traces of movement. The non-overt subject of an infinitive is PRO, while missing subjects of tensed clauses are pro. In contrast to PRO, which must be ungovemed according to the PRO-theorem, pro is found in governed position since it alternates with overt NPs. It is generally assumed that nominative case is assigned tp an NP in the specifier position of IP through government by a finite [+tense,+AGR] INFL. A non-finite [-tense,-AGR] INFL on the other hand, is not a governing head, leaving [Spec,IP] ungoverned so that PRO

occurs in that position.

(10) a. b. IP IP

PRO י pro I I VP I VP

+tcmc • ш е 4 AGR *AGR

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-17- The central concern for this study is the different manifesta- tions of subjects of finite clauses (as pro or as lexicalized pronoun) in Slavic languages, with a special focus on Russian, Czech and Upper Sorbian.

More evidence supporting the structural presence of a subject position can be drawn from the modules of binding and control theories. Consider the following Czech sentences and the binding facts they demonstrate:

'^he,) loves himself) י e miluje sebe (11)

(12) e miluje ho 1 (he^ loves him.^*

(13) e miluje Petra 1(he^ loves Peter^'

These facts follow immediately from principles A, В and

С of the binding theory, if we assume the presence of an (empty) element in [NP,IP] position. The following examples show that the theta-criterion alone cannot justify the existence of all empty subject NPs:

(14) e temneet [Ru] 1(it) is getting dark' (15) a. e prši b. e zdá se, že [Cz] '(it) is raining' 1(it) seems that'

The verbs in (14) and (15) do not assign any obvious external theta-role to the NP in subject position.6 However, there is

6 According to Chomsky (1981: 325), subjects of so- called weather-verbs have a ”quasi-argumentai" status. For some languages such a distinction of non-referential subjects

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access another (admittedly somewhat dubious) principle within GB, known as the extended projection principle (EPP), which stip- ulates that all sentences must have a subject, regardless of their argument structure. Here I follow the traditional GB analysis according to which in the sentences above a null- expletive subject occupies the structural subject position. Similarly, in overt subject languages, such as English, the EPP accounts for the presence of the expletive it or there in subject position.

(16) *в/it is nice that you came

However, some researchers, including Babby (1989), reject the EPP and instead postulate the existence of truly subjectless sentences, i.e. sentences without a structural subject position. Of course, the existence of null expletives raises obvious empirical problems. Why should they exist, (after all they are invisible and have no meaning), and how do we know and learn that they exist? One might simply want to claim that the obligatory structural presence of a subject position is part of UG. Chomsky (1982: 10) does exactly that when he writes about the EPP: may be crucial. In German, for instance, quasi-argumentai subjects, as in (ii), and non-argumental subjects, as in (i), behave differently: The latter must remain lexically unexpressed, while the former must be overt. (i) ich gehe nicht raus, weil e/*es dort getanzt wird 'I don't go outside because there is dancing going on 1 (ii) ich gehe nicht raus, weil es/*e regnet 'I don't go outside because it is raining1

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-19- We may think of it as a general principle governing D-structures, hence also governing structures derived from them. There are good reasons, which I will not elaborate here to assume that the same is true of languages that appear to allow a "missing subject'1 (e.g. Romance languages other than French) .

It is interesting that other theories also postulate the existence of such covert expletives due to theory internal requirements, such as Relational Grammar's "Final-One Law". Be this as it may, while currently we cannot really prove that these empty elements are structurally present, it is at least

as impossible to prove that they do not exist, due to their .property ״special "no meaning" and "no lexical material

2.3. Development of the pro-drop parameter Languages that leave a pronominal subject unexpressed traditionally are called pro-drop languages, since they seem to "drop" the subject pronoun. This term stems from earlier analyses and reveals that the phenomenon of null subjects was originally interpreted as a deletion rule (Perlmutter 1971).7 Thus, from the perspective of the current syntactic framework ־*the term '1pro-drop" is obsolete, since the deletion of a sub ject in the course of a derivation would contradict several

7 In earlier accounts it was claimed that the pro- drop phenomenon correlated to a host of other syntactic properties, such as free subject inversion and apparent viola- tions of the that-trace filter, suggesting that a single grammatical property might account for a whole complex of contrasts between null-subject and overt-subject languages (cf. Perlmutter 1971; Taraldsen 1978; Jaeggli 1982; Rizzi 1982) . This idea was later challenged by Safir (1985), among others.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-20-

theoretical assumptions, most notably the theta-criterion and

structure-preserving principle. However, in keeping with the

current literature on the topic, we shall maintain the label pro-drop parameter here. The development of the pro-drop pa-

rameter over the years reflects one of the most important tendencies of GB theory in general: a movement away from spe-

cific rules and towards general principles. Since Taraldsen (1978), the pro-drop parameter has been associated with the phenomenon of *1rich inflection" which can determine the grammatical features of a missing subject. He analyzed null subjects as base generated anaphoric NPs, rather them as the result of "pronoun deletion". According to him, empty subjects are null anaphors which do not violate Chomsky's nominative island constraint1 on nominative anaphors due to the fact that the null-subject NP and the finite verb are coindexed. Thus, in languages with rich agreement the subject-verb agreement can bind a null-subject NP, thereby rendering the NIC

inoperative. Taraldsen's analysis can be rephrased within a theory which now contains the ECP instead of the NIC. A subject pronoun can be omitted if AGR is rich enough to properly govern the subject position. As far as determining the type of the empty category in the subject position of tensed clauses, the investigation of null subjects predates the discovery of pro. Chomsky (1981)

8 The NIC (replaced by the ECP) stipulated that nominative anaphors have to be bound within their clause.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00062018

־21־

still assumes that there are only two types of empty categories, PRO and traces, and that the missing subject of a tensed clause is of the PRO type. As we saw earlier, PRO is charac- terized within the GB framework as a pronominal anaphor, and under that assumption it has to be ungoverned. Chomsky (1981) then has to stipulate that in null-subject languages the subject of tensed clauses may be ungoverned at S-structure due to movement of INFL into VP in the syntax. This analysis contra- diets those by Taraldsen (1978) and Rizzi (1982) f9 where the empty subject is said to be licensed precisely under government by INFL. It is not until his 1982 monograph that Chomsky posits the existence of the empty pronominal category pro, in the process of developing the full typology of empty elements. Most approaches since Taraldsen have identified the cru-

cial property of null-subject languages in the inflectional element, usually relying on the notion of richness in one way or another. One of the earliest and most influential GB ас-

counts of the null-subject parameter is given by Rizzi (1982). Perhaps the most important aspect of Rizzi 's paper is his argument that what had been considered a monolithic parameter of UG should really be split into two autonomous yet connected subparameters. His argument is based on the different distributional restrictions of referential and *1dummy" empty subjects in Italian. The first subparameter specifies whether a

Rizzi (1982) was already available as a manuscript in 1980 and Chomsky (1981) refers to it in several places.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access language licenses null subjects (INFL [♦pronominal] ) . This property is not directly related to the morphological richness of the language. The second subparameter, however, is directly related to morphological richness and determines whether INFL can recover the referential content of an empty subject (INFL [♦referential]. These two parameters predict four theoretically possible language types (Rizzi 1982: 143):

1 2 3 4 ♦pron ♦pron -pron -pron

♦ref -ref ♦ref -pron

Language type 1 is represented by Italian, a canonical pro-

drop language. Type 4 is a canonical non-null-subject Ian- диаде like English. According to Rizzi, combination 3 is excluded for "intrinsic reasons". A pronominal INFL can be referential only if its grammatical features of agreement are structured beyond a certain richness threshold. In other words, an inflection cannot be [♦ref] if it is not [♦pron]. Type 2 represents a language type which has phonetically null dummy subjects but overt referential subjects. Rizzi (1982) points out that this language type seems to exist, however, he does not go into much detail.10

10 Rizzi (1982: 143) cites the northern Italian dialect of Padua as a potential example for this parameter combination. Later research provided more evidence for this "mixed" language type (e.g. Russian, German, Icelandic).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access For any analysis of empty elements, null subjects included, there are two questions which need to be answered: What formally licenses them? How is their content (the person- number features of the missing subject) determined? Early approaches make no clear distinction between licensing and identification. Even for Rizzi (1982), who in principle argues in favor of a split of the pro-drop parameter into licensing and identification, it is INFL which licenses the null subject of a tensed clause, and it is the spelling out of the pronominal features of INFL which allows identification of the missing subject. Early work on the topic of null subjects concentrated almost exclusively on rich agreement languages, and thus it was supposed to be rich agreement alone which made pro possible. It was not until other languages were included in the research that it became apparent that the situation was far more complex.

Rizzi (1986) criticizes the standard approach under which pro is restricted to occur only in the context of rich agreement and which unifies the formal licensing of pro and the identification of its content into one and the same operation via government by a strong Agr. He points out that in the broader context of null elements such an assumption is surprising, since for traces and PRO the licensing conditions and the recovery operations are kept separate. Formal licensing appears to involve some kind of government relation, while the

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access recovery procedure involves some kind of binding relation.

Based on the occurrence of null objects (pro) in Italian despite the lack of object agreement. Rizzi (1986) argues in favor of divorcing the licensing conditions for pro from the inflectional system. He proposes what has become known as the pro-module, made up of two main independent principles (Rizzi 1986: 521,524):

(17) Pro-module: A: pro is licensed if case-marked by Xy° B: pro has the grammatical specification of the features on X coindexed with it

The class of licensing heads can vary from language to lan- диаде and seems in a sense arbitrary. In Italian, for instance, INFL. is a licensing head for pro in subject position, and V is a licensing head for pro in object position.11 Jaeggli and Safir (1989) maintain the sharp distinction between licensing and identification, however, they present a novel treatment of the two processes. Chinese and Japanese, both null-subject languages which completely lack any sort of agreement, show clearly that the licensing criterion cannot be exclusively associated with rich morphology. Jaeggli and Safir's generalization is that null subjects seem to be licensed in languages with either very rich agreement or no

1 Rizzi's (1986) study implies that pro is not restricted to subject position of tensed clauses. However, null subjects remain the main focus of this study.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access agreement at all, and they formulate morphological uniformity as a licensing criterion.12 An inflectional paradigm of a language is morphologically uniform if it has either only underived inflectional forms or only derived inflectional forms. If some forms of a paradigm are divisible into stem+affix and other forms of the same paradigm are bare stems, then the paradigm is mixed, and no null subjects are licensed. Under this model, the licensing conditions do not provide a mechanism to recover the referential value of a null pronoun. A separate identification mechanism is necessary to determine the content of a thematic NP. According to Jaeggli and Safir, identification can come about in a couple of dif-

ferent ways. However, they are fairly vague on formulating the precise conditions for identification. Identification can happen locally by agreement if the category containing AGR case-governs the empty category.15 Thus, in languages like

12 The discovery that the phenomenon of "pro-drop" can occur when there is a rich agreement element or when there is no agreement at all actually goes back to Huang (1984). 13 It has been argued that in V-2 languages, such as German or Icelandic, the nominative case is assigned by Tense, which is located in COMP (C of CP), while AGR is located in INFL (I of IP) . Under the requirement that identification by AGR can only happen if the category containing AGR is case-governing the empty subject, we now have an account for the lack of thematic null subjects in German and Icelandic. Both languages have morphologically uniform paradigms, and hence license null subjects, however, identification is blocked due to the distribution of Tense and AGR in separate nodes. More support for this approach comes from West Flemish (cf. Bennis and Haegemann (1984), which has complementizers which are inflected for person and number and indeed allows null thematic subjects. This is, then, an instance in which both Tense and Agr are part of the same governing node, and identification can take place.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Spanish or Italian, the agreement affixes with the relevant phi-features act as identifiers. For languages which uniformly lack agreement, such as Chinese and Japanese, Jaeggli and Safir essentially follow Huang's (1984) analysis. In these languages, a null subject can be identified either by a c- commanding nominal in the matrix clause or by a (null) topic.14 Although it is not clear per se why something like the morphological uniformity criterion should be operative in

1 Only in the first instance should we really speak of pro. Consequently, the occurrence of genuine null subjects in Chinese and Japanese is in fact quite restricted. Consider the following Chinese examples (Huang 1984): Speaker A: Zhangsan kanjian Lisi le ma? 1did Zhangsan see Lisi' Speaker B: (i) в kanjian ta le 1(he) saw him1 (ii) Zhangsan xiwang [e keyi kanjian Lisi 'Zh. hopes (he) can see Lisi1 In Huang'8 analysis, the empty element in (i) can only be a variable bound to a zero topic. In "rich" agreement languages which license pro the empty element in the same position could be identified by Agr. This type of identification is unavailable in Chinese. The empty category in (ii) may have dual status. Its reference can be determined by an "outside" discourse topic (variable reading) distinct from Zhangsan or it can be fixed via coindexation with the NP Zhangsan in the matrix clause (pronominal reading). One might object that even in a pro-drop language the reference of an empty third-person subject, such as in (i), seems to depend on the discourse. The difference, however, is the following: The interpretation of 3rd person pro could be understood as a two-step process. First, the pronominal features under AGR determine the pronominal content (person, number, gender) of the empty pronoun. This is what I refer to as identification of the content of pro. Once that is done, the identified pro behaves like a "normal" pronoun, being interpreted according to principle В of the binding theory. During that part of the interpretation operation a pronoun may be freely coindexed with NPs outside their governing domain.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-27- natural language (cf. Speas (1994) below for a possible interpretation), the licensing and identification theory developed by Jaeggli and Safir fits well into the general context of empty categories. Similar to Rizzifs (1986) pro-module, it keeps licensing and recovery of content separate. Licensing is a purely formal criterion, while identification is a more contentful criterion, referring to the availability of some mechanism to retrieve the essential grammatical information left phonologically unexpressed in order to provide the empty category with some content. For traces, this division of labor is exemplified by the ECP as the licensing condition, and antecedent government as the identification operation. The empty category PRO is limited to occur in ungoverned con-

texts, and derives its content from an antecedent which is determined through control theory. Originally it was assumed that there were only two Ian- диаде types with respect to the pro-drop parameter: Null-subject languages (Italian type) and non-NSL (English type). However, as discussed earlier in this chapter, Rizzi (1982) predicted the existence of a "mixed1* language type. German, is one such semi-pro-drop language. It does not omit thematic subject pronouns, as shown in (18), though it does permit (requires) expletive subjects to be null, as shown in (19).15

15 In fact, there are probably two types of semi-pro- drop languages, namely those which allow only non-argumental subjects (excluding subjects of so-called weather verbs which are quasi-argumental according to Chomsky 1981), for instance German, and those that allow both non-argumental and quasi- Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access (18) er/*pro tanzt gern 'he likes to dance1

(19) P geht ins Dorf, weil pro/*es dort getanzt wird 'P goes to the village, because pro there is dancing'

A licensing/identification approach à la Jaeggli/Safir predicts the difference between the German type (semi-pro- drop) and Italian type (full pro-drop). It makes sense that within such a scheme, empty categories which lack semantic content need to meet only the licensing criterion, since there is no content which needs to be recovered via identification. Expletive subjects, by definition, lack any semantic content, and therefore only need to be licensed. In such a system then, the distribution of null-subject types in German can be understood as a result of formally licensing but not identifying null subjects. The licensing/identification theory might be understood as breaking up the originally monolithic pro-drop parameter into two distinct parameters, which can be paraphrased as a "Null-Referential-Pronoun" (NR) parameter and a "Null-Exple- argumentai subjects, for instance Icelandic and Yiddish. Unless otherwise stated, for our purposes both non-argumental and quasi-argumentai subjects are subsumed under the term -Rizzi (1986: 541) lists the following possi .״expletives" bilities: a : no occurrence of pro b: pro=non-argument (German) c: pro*non-argument and quasi-argument (Yiddish, Icelandic) d: pro=non-argument, quasi-argument, and referential argument (Italian) Compare the following examples from Icelandic and German: (i) hefur pro rignid (ii) hat es/*pro geregnet

*has (it) rained' 'has it rained'Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access tive" (NE) parameter. However, as already implied in Rizzi (1982), these two parameters are not completely independent, i.e. not all four theoretically possible combinations of the two binary parameters should actually exist.16 A combination of [♦NR,-NE], that is a language with null referential and overt expletives, is unexpected. Expletives by definition lack any referential function, so the only reason for their occurrence must be that a language does not license null subjects at all. In other words, it is much easier for an expie- tive to be null than for a referential pronoun to be null, since there is no pronominal content which must be recovered.17 There are two main problems with the licensing condition as formulated by Rizzi (1986) and the morphological uniformity hypothesis: First, neither encodes the longstanding intuition that there is some sort of connection between rich agreement and the occurrence of null subjects, and second, it is not clear why languages like Italian and Japanese pattern together with respect to null-subject phenomena (if they really do) . Although the remainder of this study will focus on the appli-

cability of the licensing/identification approach to Slavic, it is worth summarizing a recent proposal by Speas (1994)

10 See pages 22-23 for a discussion of Rizzi (1982). 17 Within this framework it makes no sense to talk about identification without licensing, since, if null subjects in some language were identified without being licensed, they still could not exist, so that the fact that they were identified would have no consequences.Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-30- which attempts to remedy these shortcomings. As theoretical background, Speas assumes familiarity with Pollock's (1989) split INFL hypothesis according to which INFL is decomposed into (at least) two separate functional heads AGR and T, each with its own phrasal projection. This yields the clause structure in (20b).IS

(20a) (20b) CP CP Spec C' Spec C' / С IP AGRP *

* 9 NP^ I ' AGRNP

Ф I VP TPAGR Spec v־ VP

Speas's (1994) proposal aims at explaining Jaeggli and Safir's generalization. Her claim is that the null-subject parameter follows directly from the principle of economy, according to which a projection XP can be projected only if it has content. A node has content iff it dominates a distinct phonological matrix or a distinct semantic matrix. Null arguments occur wherever general principles of economy permit them,

!в I am noncommittal as to the respective position of AGRP and TP, since this has no bearing on the current study. In fact their position might be subject to cross-linguistic variation. The theory of functional projections has received much attention recently, and several other functional projections have been postulated, such as Negation Phrase, Focus Phrase and Determiner Phrase. Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access and no special licensing conditions on pro are necessary. Adapting a proposal by Rohrbacher (1995), Speas proposes that in languages with "strong" agreement, each agreement morpheme has its own lexical entry, while in languages with "weak" agreement, the morphemes do not have independent entries. Verbs in weak agreement languages are listed in the lexicon in verbal paradigms. According to Rohrbacher (1995: 356) "strong" agreement, and therefore lexically listed affixes, are found

only in those languages which meet the following conditions.

(21) Conditions for lexically listed affixes: AgrS is a referential category with lexically listed affixes in exactly those languages where regular subject-verb agreement minimally distinctively marks all referential AgrS-features such that in at least one number of one tense, the person features [1st] and [2nd] are distinctively marked.

The functional head AGR has no semantic content. There- fore, if projected it must have phonetic content. In languages with strong agreement, an agreement morpheme heads its own functional projection AGRP. Such an AGRP has content without its Spec position being filled by an overt NP. In weak AGR languages, on the other hand, the affix is base-generated on the verb, and so [Spec, AGRP] has to be filled in order to give AgrP content. The former situation accounts for null- subject languages, the latter for overt subject languages. The principle of economy of derivation also accounts for the Asian null-subject type. Speas suggests that the AGRP projection is necessary only in languages with some sort of agree- Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־32־ ment. In languages with no agreement whatsoever, there is no AGRP at any level, therefore the language has null subjects since nothing forces movement into the specifier of AGRP.19 For Speas, there is no null-subject parameter per se; instead,

null-subject properties of a language are accounted for by an ״AGR strength parameter coupled with general principles of eco nomy of representation.

2.4. Hull-subject diagnostics Ideally, we would have only to look at a sentence involving a pronominal subject in order to determine whether or not a particular language belongs to the null-subject type. However, many (if not all) languages can omit various elements, including subjects, in various styles of discourse. English certainly should not be analyzed as a null-subject language. Nevertheless we find English sentences without overt subjects, for instance when recounting a sequence of events in "diary" style or in certain questions.

(22) a. Woke up. Fell out of Bed. Dragged a comb across my head. b. Got milk?

Similarly, in German a pronoun can be dropped under certain contextual conditions, provided it occupies the initial

19 It is not clear whether it makes sense to speak of "null subjects" in this context at all, since no structural subject position [Spec,AGRP] is available.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־33־ position in a root clause, or, in other words, the [Spec,CP] position (cf. Ross 1982 as cited in Huang 1984: 555/56).

(23) A: Weißt du schon, daß Peter wieder zurück ist? 1do you know already that P. is back again1 B: Ja, ja. e Hab ihn schon gesehen. 'yes,yes. e have already seen him1

The discourse related omissibility of elements, called ellipsis, varies from language to language, and defining the exact conditions under which such '1loss'1 of lexical material can happen lies outside the scope of this work.20 In some lan-

guages, however, including Russian, missing subjects due to ellipsis can occur in such a wide range of conditions that it becomes difficult to differentiate between genuine pro-drop and ellipsis.2*

It follows that one cannot readily determine the status of missing subjects in a language, and it has been suggested

that we should instead rely on a set of syntactic tests that can be used as diagnostics of pro and therefore the null-sub-

There are several interesting questions which I do not plan to discuss here: Should we posit an empty category in the place of the deleted element? If so, which type of empty category should be posited? For a very good discussion of the issues involved see Huang (1984). He analyzes the missing pronoun as a variable in argument position bound by a zero topic, i.e. referring to the discourse topic. The zero topic in turn is discourse-bound to the initial topic of a topic chain. Huang clearly differentiates genuine zero pronouns and variables bound by zero topics. He proposes the existence of two independent parameters: [!zero-topic] and [±pro-drop] . with four logically possible types of language. 21 For a more detailed discussion see chapter 3.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-34- ject status of a particular language (Lindseth and Franks (1994); Franks (1994)).

Typical characteristics of null-subject languages are ע.(given in (24

(24) a. Only null pronominais are stylistically unmarked.

b. Only null pronouns can function as bound variables.

c. Only null 3rd plural pronominal subjects can have arbitrary reference.

(24a) expresses the well-known fact that поп-emphatic pronominal subjects are regularly omitted in canonical null-subject languages, while overt pronominal subjects are emphatic. Criterion (24b) has been adopted from Montalbelli (1984), whose analysis of pronominal binding in null-subject languages revealed that empty pronouns do not behave in the same manner as their overt counterparts. In particular, he discovered

that there is a curious asymmetry between the interpretation of overt and empty pronouns. Montalbelli (1984) claims that in the following type of construction in a null-subject lan- диаде an overt pronoun in the same position as pro cannot be interpreted as bound, whereas the empty pronoun has to be interpreted as bound:

22 Technically speaking, we have to differentiate between two types of null subject languages: a) languages which license and identify null subjects b) language which just license null subjects In the following I use the term (canonical) 'null-subject language1 synonymously with type a), unless otherwise stated.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-35- (25) a. muchos estudiantes piensan que ellos son inteligentes ,many students think that they are smart' b. muchos estudiantes piensan que pro son inteligentes

I do not completely agree with Montalbelli that the only interpretation which is possible for (25b) is a bound reading:

(26) (many x: x a student) x thinks x is intelligent

Instead, a referential (free) interpretation of the empty pronoun is also possible, even though this is not the most likely interpretation, especially if a sentence is given to a

native speaker without any context. However, what is most crucial for our discussion, the bound variable reading in (26) is not available for (25a). In English, on the other hand, the interpretation of the pronoun they is ambiguous between a bound and free interpretation. Montalbelli emphasizes that overt pronoun constraint" is only valid for positions״ his where the overt/empty alternation is available. The interpretation of the overt possessive pronoun in sentence (27) is ambiguous in both Spanish and English. In Montalbelli1s account this is expected, since the overt possessive pronoun does not have an empty counterpart.

(27) muchos estudiantes piensan que sus bicicletas son azules ,many students think that their bicycles are blue'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-36- A version of Montalbelli's overt pronoun constraint can then be defined as follows:

(28) Overt pronouns cannot be bound if the option overt/empty is available.

Criterion (24c) is derived from Jaeggli*s (1986) analysis of the so-called arbitrary plural construction. Jaeggli points out that this type of construction occurs in both English and Spanish, indicating that its existence is independent of the setting of the null-subject parameter. However, while in English it is the overt pronoun they which functions as the arbitrary plural pronoun, as illustrated in (29), in Spanish only a null pronoun can be interpreted as arbitrary in reference. According to Jaeggli (1986) an

overt pronoun in Spanish can only be interpreted as non-arbitrary. In (29) the lexical pronoun they is ambiguous between a specific referential reading (a group of more than one person) and an arbitrary reading (someone; one or more people). In Spanish, there is again a division of labor between overt and null pronouns, although perhaps less clearly so than stated by Jaeggli. Examples (30a) and (30b) below are cited from Jaeggli (1986). Some Spanish informants were less willing to

interpret (30b) with an arbitrary reading, but all agreed that

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-37-

(30a) certainly cannot have an arbitrary reading. The same ע.(results obtained in the constructed examples (31a) and (31b

(29) they don't allow alcohol on campus (30) a. ellos 11aman a la puerta 1they calljpi at the door1

b. pro llaman a la puerta (31) a. aqui ellos no permiten las bebidas alcohólicas 'here they don't allow alcohol' b. aqui pro no permiten las bebidas alcohólicas

The pronoun ellos in (30a) and (31a) must have a specific group of people as reference, while the null pronominal in (30b) and (31b) is ambiguous between definite reference and

arbitrary reference. Jaeggli unifies this property of null- subject languages with Montalbelli's analysis, proposing that the overt pronoun constraint and the constraint which does not allow an overt pronoun to have arbitrary interpretation in null-subject languages, are reflecting the same underlying

principle. In contexts where the overt/empty alternation is available, overt pronouns must be specified as [-dependent], that is independent in reference, while empty pronouns are [♦dependent], fixing their reference by linking to an antecedent

Part of the problem seems to be that Spanish has מ another tensed construction with arbitrary null subject, the so-called impersonal se construction, which native speakers tend to prefer over the arbitrary plural construction discussed here.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access through some sort of binding mechanism.24 As for the arbitrary plural construction, a [♦dependent] third-person pronom- inai is assigned the arb index at LF if it lacks an antecedent. Jaeggli's constraint on the arbitrary interpretation of overt pronouns is then defined as follows:

(32) Overt pronouns may not be arbitrary in reference iff the overt/empty alternation obtains.

This generalization raises various problems, the solution to which lies outside the scope of this work. Jaeggli is not very specific as to what kind of binding relation "in an extended sense of the term" is involved for [+dependent] elements to fix their reference. For instance, in Spanish any null pronominal in subject position is specified [♦dependent]. Thus, in the following sentence the null pronominal has to fix its reference by linking to an antecedent, supposedly by being bound by AGR located in the finite INFL. (i) pro compro un pulpo 1he/she bought an octopus' AGR in this instance contains only person and number features. Therefore, when uttered outside any context the sentence is ambiguous between he and she, so that coindexation with some "outside" antecedent seems necessary to fix refer- enee to gender. Furthermore, [-dependent] pronouns also seem to fix their reference via "some kind of binding relation" to an antecedent. Also consider the problem which arises in the following type of constructions: (example from Montalbelli 1986: 140) (ii) muchos carniceros creen que pro compraron un pulpo 'many butchers think that pro bought an octopus' Here the [♦dependent] pronominal subject has to act as a bound variable, being bound by ,many butchers'. However, it seems that at the same time it will be bound by the AGR of its own clause, for the same reasons as in the above example (i). A third connected question is whether or not empty 1st and 2nd person subject pronouns in null subject languages should also have the specification [♦dependent] or not. Note that for them, being bound by INFL does suffice to fix their reference.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-39- While we are not concerned about the details of Jaeggli's

analysis, it is important because it constitutes the basis for the third of our null-subject diagnostics. In summary, the characteristics under (24) can be taken as fairly reliable diagnostics as to whether or not a language belongs to the canonical null-subject languages. In the next chapter these diagnostics will be applied to Slavic.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 3. NULL-SUBJECT CHARACTERISTICS OF SLAVIC LANGUAGES 3*1• Application of null-subject diagnostics The last section of the previous chapter presented a set of three diagnostics (repeated here as (1)) to determine the canonical null-subject status of a language.

(1) The following properties hold true for null-subject languages :

a. Only null pronominal subjects are stylistically unmarked. b. Only null pronouns can function as bound variable. c. Only null 3rd plural pronominal subjects can have arbitrary reference.

The next step is to apply these diagnostics to Slavic. We shall take Russian as representative for East Slavic, Serbo-

Croatian for South Slavic, and Czech for West Slavic.

3.1.1• W«at Slavic and South Slavic By these criteria most West and South Slavic languages are canonical null-subject languages.1 Both Czech and Serbo- Croatian do not normally lexicalize pronominal subjects of tensed clauses, as illustrated in (2) and (3) . The use of overt pronouns is generally reserved for contrast, emphasis, or a change in discourse topic, shown in (4) and (5).

For Sorbian, which does not behave like a canonical null-subject language, see Chapter 5.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access (2) öasto pro vstávám brzo [CZ] (3) öesto pro ustajem rano [SC] 'often (I) get up early1

(4) Alena vstává brzo, a já často vstávám pozdë [CZ] (5) Jovan ustaje rano, a ja Casto ustajem kasno [SC] 'A/J gets up early, but I often get up late'

Diagnostics (lb) and (lc) provide further evidence that Czech and Serbo-Croatian are null-subject languages. The Czech examples under (6) and Serbo-Croatian examples under (7) pertain to (lb), demonstrating that only null pronouns can have bound variable readings.

(6) a. každy student si mysli, že pro dostane jedničku b. každy student si myśli, 2e on dostane jedničku (7) a. svaki student misli da će pro dobiti desetku b. svaki student misli da će on dobiti desetku 1every student thinks that he will get an A 1

The bound variable reading rendering interpretation (8) can only obtain in examples (6a) and (7a).2

(8) ((every x: x a student) x thinks x will get an A)

The fact that the overt pronoun in the example below may be bound to the quantifier phrase rendering inter- prêtât ion (ii) is not a contradiction to what has been said so far. In the position after a preposition there is no alternation between empty and overt pronouns and hence Montalbelli's overt pronoun constraint does not apply. (i) každi student chce, aby Maria hrála s nim (ii) ((every x: x a student) x want that M. play with x) Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00062018

-42- Examples (6b) and (7b), which have overt pronouns, can only be understood in the sense that every student thinks that a particular person is getting an A. Examples (9) and (10) provide an application of diagnostic (1c). In Czech and Serbo-Croatian, the arbitrary reading obtains only if the subject is 3rd person pro instead of overt oni.

(9) a. pro v tom obchodë prodávají kávu [CZ] b. oni v tom obchodë prodávají kávu ' in this store they sell coffee'

(10) a. pro ovdje prodavajи kavu [SC] b. oni ovdje prodavaju kavu *they sell coffee here'

The overt pronouns in (9b) and (10b) necessarily give rise to the [-dependent] interpretation, so that oni in those sentences must refer to a specific group of people.

3.1.2. Russian (Bast Slavic) It is easy to find Russian examples with zero subjects in tensed clauses, especially in colloquial Russian or artistic texts imitating colloquial style. The following examples were found in stories by Sokolov, Jerofeev, Romanina (Innis 1985) .

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Missing subjects in the Russian text are marked with #, other missing material is given in brackets.3

(11) On zamečal, čto # dumáét о nej postojanno, i radovalsja, čto # ničego ne xoãet ot neë, i značit # dejstvitel'no ljubit (её). 'Не realized that (he) was constantly thinking about her, and rejoiced that (he) didn't want anything from her, that means (he) really loved her. ' (12) No daže esli by ja pomnil sejčas её imja, to # ne stal by nazyvat' (ego), какое vám deio. 1But even if I remembered her name right now, then (ï) would not name it, what business of yours is ״ .it (13) Ja ej i stēkla skol'ko raz vstavljal, i peč1 perekladyval,. . . a ni razu # ne ugostila (menja). Den'gi # daët (mnę), a nasëèt ètogo noi1. 1I put windows in several times for her, and moved the stove, but (she) didn't invite me for dinner a single time. (She) gave me money, but nothing ever came of it.' (14) Budet vremja - i kryäu # počinju. A v tot den1 s utra stêkla # vstavljaju. 'If there is time - (I) will fix the roof as well. But on that day (I) I'm putting in windows from early morning on.1

Non-overt subjects are very frequent when clauses are coordinated: (i) on vošēl v komnatu, # sei za stoi i # načal rabotat ' 'he entered the room, sat at the table and started working' I analyze these cases as VP conjunction, and therefore not as instances of empty/missing subjects. The second and third verbs "share" the subject with the first verb. Therefore, sentence (i) should be represented by the structure (ii) and not (iii). (ii) [$[NP°n] [vptvp vošėl v komnatu] i [Vp sei za stol]] (iii) [ [s [NP on] [Vp vošel v komantu] ] i [s e [Vp sei za stol]]] Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access (15) Prokuror - on pobelel ves', # govorju daže gazetu nogami ־ perestal čitat1. # Xodit po učastku oduvančiki # topčet. 'The state attorney turned completely pale, (I) say, and even stopped reading the newspaper. (He) walks along the plot, with his feet (he) tramples dandelions.' (16) Vse govorjat: Kreml1, Kreml*. Oto vsex ja slyšal pro nego, a sam # ni razu ne videi (ego). Skol'ko raz uže, tysjaőu raz, napivâis' ili s poxmeljugi, # proxodil po Moskve s severa na jug, s zapada na vostok ... - i # ni razu ne videi Kremlja. •Everybody says: The Kremlin, the Kremlin. From everybody I hear about it, but (I) myself haven't seen it a single time. How many times already, a thousand times, drunk or with a hangover, did (I) walk through Moscow from North to South, from West to East, but not a single time did (I) see the Kremlin.1

Several authors (Müller (1989); Kosta (1990); Benedicto (1994)) have argued that Russian is a canonical null-subject language. One reason for this judgement is undoubtedly the wide range of examples with missing subjects which can be cited. Below I will argue against this claim on the basis of the null-subject diagnostic tests. I essentially agree with Franks (1990), who analyzes missing subjects in examples comparable to the ones above as being deleted at PF due to some kind of discourse-licensed ellipsis. Such loss of lexical material in Russian is clearly not restricted to subjects, and thus should be treated separately from the pro-drop phenomenon. (Colloquial) Russian in general has a strong tendency toward the deletion of lexical material which is recoverable from the context, so that the "loss" of subject pronouns is

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access just a part of this larger phenomenon.4 Examples (11-13), and (16) cited above illustrate deletion of direct and indirect objects. Ellipses of complementizers as shown in (20) and (21) (examples cited from Nakhmovsky and Leed (1988)) and verbs as shown in (17) and (18) are also frequent:5

Huang (1984) proposed a "discourse-oriented vs. sentence-oriented" parameter. Nichols (1985) suggested that Russian is a very discourse-oriented language. Discourse- oriented languages tend to follow a rule according to which topic NPs are deleted under identity with a topic in the preceding sentence. Example (13), repeated below, illustrates the complexity of this phenomenon:

(i) Ja ej i stëkla skol'ko raz vstavljal, i peč ' perekladyval,.. a ni razu (ona) ne ugostila (menja). Den*gi (ona) daêt (mne), a nasčēt ètogo noi'. 1I put windows in several times for her, and moved the stove, but (she) didn't invite (me) for dinner a single time. (She) gave (me) money but nothing ever came of it.1 The speaker and a female person are the "known entities" or topics of the discourse. The omission of the first ona is particularly interesting, since this is actually a case of contrast, where we would expect overt pronouns in a pro-drop language: 'I did this, but she didn't do that...' It is also worth mentioning that most Russian speakers, while allowing missing objects due to ellipsis, do not allow "syntactically active" pro-objects, as in Rizzi (1986): (ii) il bel tempo invoglia a restare ' the good weather invites pro [PRO to stay] ' *xoroãaja pogoda soblaznjaet ostavat'sja 5 Similar elliptical constructions with missing verbs or complementizers were rejected by Serbo-Croatian informants. Example (i) equates to (ii) in Russian as (iii) does to (20) . (i) *Knjigu # vam? (dati) *Should I (give) you the book' (ii) knigu # vam? (iii) *Dima došla # Jovan još spavao. (kada) This indicates that there is not a direct relationship between the ellipsis phenomenon and pro-drop properties of a language.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access (17) kuda ty (idēš') 'where do you (go)1 (18) ty (prixodiS1) ko mne 'do you (come) to me' (19) Dima prišēl (kogda) ja eščē ležala 1Dima came (when) I still was lying1 (20) u nas novaja sosedka (kotoruju) Natašej zovut 'we have a new neighbor (whom) they call Natasha'

Here I will not be concerned about exactly which factors allow recoverable lexical material to be deleted at PF, nor

will I attempt to determine if a "bound variable and zero topic" analysis à la Huang (1984), as discussed in the previous chapter, would also apply to Russian.6 The real problem for this work is that (what are in my view) discourse licensed null subjects may appear to exhibit the null-subject properties in (1), giving the (false) impression that Russian is a canonical null-subject language.7

At first glance it might be tempting to attribute ellipsis of subjects to "rich" INFL. However, this ,'explanation" fails to account for the deletion of objects, verbs and conjunctions, and once we claim that zero subjects in Russian are part of a larger phenomenon we would also like to give a unified account of the recoverability conditions of deleted constituents.

7 One might actually want to argue that missing subjects in Russian are not stylistically unmarked, hence fail- ing criterion (la). Instead, they may be marked, perhaps as "intentionally colloquial". Jakobson (1939 :150) wrote in that context: " ... en russe c'est justement l'omission excessive du pronom de la première personne que Dostojevski éprouve comme une morgue irritante" ('... in Russian it is precisely the excessive omission of the pronoun of the first person which Dostoyevsky felt to be an annoying arrogance') .

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-47-

Since the diagnostic tests boil down to a division of labor between overt and null subjects in null-subject languages, one can couch the same conditions in terms of overt subjects.8

(l') a. Overt subjects are stylistically marked.

b. Overt subjects cannot function as bound variables. c. Overt subjects cannot be arbitrary in reference.

The important point is that overt subjects in Russian also exhibit the properties under (1) and not those under (l1).

This means that Russian does not exhibit the division of labor between null and overt subjects which one would expect for a

true null-subject language. In order to avoid misjudgments based on the occurrence of null subjects in languages like Russian, the following investigation will focus on the interpretive possibilities associated with overt subjects as summarized above in (11 ) .

First, in contrast to most West and South Slavic languages as exemplified by Czech and Serbo-Croatian, there is nothing

emphatic about the overt expression of a pronominal subject in Russian.

(21) a. ja ponimaju vas [Ru] b. já vas rozumím [Cz] c. ja vas razumem [SC] 1I understand you1

See Lindseth and Franks (1994).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־48■ In Czech and Serbo-Croatian, the presence of an overt subject clearly indicates opposition, contrast or emphasis, while this is not the case for Russian. In neutral literary style, the use of overt pronouns in Russian is often obligatory, and the dropping of a subject is felt as incomplete and elliptical by native speakers. The normal way of saying 'I love Peter' in Russian is ja ljublju Petra with the subject ja, while in Czech the normal way is milu je Petra without a subject. Rus- sian speakers learning Czech or Serbo-Croatian find the regular lack of pronominal subjects very striking and must make a

conscious effort to avoid the overuse of overt pronouns. That should not be the case if Russian were a canonical null-subject language. Sentences (22-24) provide more examples of this contrast between Czech and Russian:

(22) a. # mluvil jste s jeho ženou [CZ] b. vy govorili s ego Ženoj [RU] ,you talked to his wife'

(23) a. # došla pro mléko [CZ] b. ona sxodila za molokom [RU] •she went for milk' (24) a. # bydlím v Americe [CZ] b. ja živu v Amerike [RU] 'I live in America'

Criterion (lb') similarly divides the Slavic languages into East, on the one hand, and West and South on the other. Russian in (25) allows the bound variable reading despite the

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access presence of the overt pronominal subject, but this option does not exist for the otherwise identical Czech (6b) and Serbo- Croatian (7b) examples repeated below as (26) and (27) .

(25) každy student dumaet, čto on polučit pjatërku [Ru] 'every student thinks that he will get an A' (26) každy student si myslí, že on dostane jedničku [CZ] (27) svaki student misli da će on dobiti desetku [SC]

In Serbo-Croatian and Czech the overt pronoun cannot be bound by the quantified expression, while in Russian (and English, for that matter), the overt pronoun can render the bound variable interpretation in (8) repeated here as (28).

(28) ((every x: x a student) x thinks x will get an A)

Roughly the same results obtain if the QP is replaced by a referential NP, as in (29-31).

(29) Ivanį dumaet, čto оіц poličit pjatërku [Ru] 1Ivan thinks that he will get an A' (30) Janį si mysli, že on.^ dostane jedničku [CZ] (31) Jovanį misli da će on™^ dobiti desetku [SC]

Using the overt pronoun instead of pro in a null-subject lan- диаде therefore leads to a deictic, i.e. non-dependent interpretation, so that the intentional coreference of an overt pronoun in a subordinate clause with the matrix subject is

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access ■50־ impossible, or at best extremely marginal, in null-subject languages.

According to the third criterion, an overt subject pronoun in a null-subject language cannot be arbitrary in refer- enee. It turns out that this test, although less obvious, further supports the proposed division of Slavic languages.9 The Russian arbitrary 3rd plural construction is normally described as requiring that the subject be phonologically null. However, some Russian native speakers nevertheless confirmed that the examples in (32) can have an arbitrary interpretation.10

(32) a. v Amerike oni govorjat po-anglijski ,in America they speak English1 b. vo Francii oni edjat ulitok 1in France they eat snails1 c. na benzokolonkax oni prodajut sigarety 1at gas stations they sell cigarettes'

Although the constructions may not be normative with respect to literary Russian, the point is that the presence of an overt pronoun does not actually force a поп-arbitrary inter- prêtâtion. In Serbo-Croatian and Czech, on the other hand.

In general, this test seems less reliable than the other two null-subject criteria. Native speaker judgements about the possibility or lack of an arbitrary interpretation of an overt pronoun vary, not only in Russian but also in Sorbian, as illustrated in Chapter 4. 10 See Lindseth and Franks 1994; Sperling 1990.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access the presence of an overt pronoun does not allow an arbitrary interpretation, as discussed above.

I therefore conclude that Russian (East Slavic) is not a canonical null-subject language, while Czech (West Slavic) and

Serbo-Croatian (South Slavic) are. The observation that Rus- sian is different from other Slavic languages with respect to the expression of subject pronominal is not a new one. Jakobson, for instance, wrote (1939: 150):

Le russe présente deux variantes stylistiques pour dire "je vais /en voiture/": ja jedu /avec pronom personnel/ et jedu /sans pronom/. De même, en tchèque: já jedu et jedu. Cependant, il y a là une grande différence entre les deux langues: ..... en russe, c'est la construction à deux termes qui est le type "normal", tandis que la variante à sujet- zéro est un procédé expressif. En tchèque, au con- traire, le zéro d'expressivité se rapporte au sujet- zéro, et la valeur expressive s'attache au type já jedu.11

Interestingly, even some authors who claim that Russian is a pro-drop language comparable to Czech do point out that there is a difference between the languages concerning the properties of zero and overt pronouns, respectively. Koktová (1992: 40,41,60), for instance, investigating the variation among "anaphoric expressions" in Czech and Russian sentences

Russian has two stylistic variants to say I'm' ״ going / by car /": ja jedu /with a personal pronoun/ and jedu /without pronoun/. The same, in Czech: já jedu and jedu. However, there is a big difference between the two languages: . . . in Russian it is the two-termed construction which is the "normal" type, while the variant with the zero-subject is an expressive mode. In Czech, on the other hand, zero expres- siveness is related to the zero-subject, and the expressive value is attached to the type ja jedu.1

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access of the type Karel, fekl. že Oj pčijde {'Karėk said that (he) would come') / Miša, skazał, čto (О^оп^ priedet ('Miša said that (he) /he would come1) writes:12

My major claim is that in Czech, the basic variation is between the zero form and the anaphoric NP as unmarked anaphoric expressions (the personal pronoun on being highly marked, occurring only in emphatic (emotional contexts) ... On the other hand, in Rus- sian, the basic variation is between the zero form or the personal pronoun on ... Unlike Czech, the relation of the primary (zero) and secondary anaphoric expressions in Russian is rather loose; these two anaphoric expressions (the zero form and the pronoun on) are almost freely interchangeable.

"Anaphoric" in this context refers to a pronoun or full NP which derives its reference by virtue of being linked to an NP in a preceding clause. The use of an overt ("secondary") anaphoric expression in a pro-drop language means that a certain amount of explicitness is necessary to avoid ambiguity. According to Koktová, the major condition for the usage of an anaphoric NP (instead of the primary anaphoric expression which is zero) in Czech is a nonsubject antecedent, especially if there is a danger of ambiguity (sentence (i) is a slightly changed version of Koktová*s example) . (i) Pavel! Alenëj nic nedal, protože Alena, zlobí 'P. didn't give A. anything, because A. is bad' The use of an overt pronoun ona would presumably be highly marked and only possible in emphatic (emotionally colored) context. Thus, the Czech and Russian examples below are not equivalent. Instead, (iii) corresponds to example (i) above. (ii) Pavel Alenë nic nedal, protože ona zlobí [CZ] (iii) Pavel Alene ničego ne dal, potomu čto ona ploxo sebja vedét [RU]

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 3.2. Why is Russian different?

3.2.1. The intuitive explanation The conclusion that Russian (East Slavic) does not qualify as a canonical null-subject language, while Czech (West Slavic) and Serbo-Croatian (South Slavic) do, leads inevitably to the question of why East Slavic is different from the other languages. First of all, the question arises whether or not Russian does not license any null subjects. Within Jaeggli and Safir's framework, this would be unexpected. The Russian verbal paradigm has only derived inflectional forms and thus seems to meet the morphological uniformity hypothesis. There-

fore null subjects should be licensed in Russian. Indeed, expletives in Russian may (and generally must) remain phonologically empty.13 None of the verbs in (33-36) assigns a theta-role to its subject. The first two are weather-verbs, (35) contains a so-called raising verb and (36) a passive.

(33) pro temneet getting-dark 'it's getting dark' (34) zdes' pro duet here blows 'there is a draft here'

13 Some dialects allow overt expletives, as can be seen in the following example taken from Woolhiser (1994): ono tëplo v izbe it warm in cabin 'it's warm in the cabin' Overt expletives in Slavic languages are discussed in Ch. 5.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access (35) mne pro kažetsja, čto on ne ponimaet menj a me seems that he not understand me ,it seems to me, that he doesn't understand me* (36) pro było napisano ob ètom v gazete was written on this in paper *They wrote about this in the newspaper'

The problem which arises is how we can determine what it is which blocks the identification of referential null subjects in Russian. Before trying to formalize this though, it is worth looking at the "old" intuition that there is a correlation between the richness of verbal agreement morphology and availability of null subjects in a language. We should be able to find a relevant difference in the overt expression of agreement features between Russian and the languages which represent West and South Slavic in this discussion. The present tense paradigm of Russian actually has distinct endings for every person/number combination, as does Serbo-Croatian which is a canonical null-subject language. Czech even has a syncretism between the 3rd person singular and plural within one conjugational pattern.14 The conjugational paradigms for verbs with infinitival a- and i- stems given in Tables 1 and 2 below illustrate this.

1 For a discussion on the allowable numbers and types of syncretisms see section 4.5.1.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Table I: Conjugational Paradigms: a-stems

RU: obedat1 CZ: obëdvat SC: ruCati lsg obedaju obëdvàm ruSam 2sg obedaeš1 obêdváá ručaŠ 3sg obedaet obëdvà ruča lpl obedaem obëdvâme ručamo 2pl obedaete obëdvâte ručate 3pl obedajut obëdvaj í ruČaju

Table 2: Conjugational Paradigms: i-stems

RU: govorit' CZ: hovofit SC: govoriti

lsg govorju hovofim govorim 2sg govoriš* hovofiš govoriš 3sg govorit hovofi govori lpl govorim hovofíme govorimo 2pl govorite hovofite govorite 3pl govorjat hovofí govore

Therefore, the present tense paradigm of Russian does not give us the desired answer. At first glance, Russian appears even more qualified to license null subjects than Serbo-Croatian. Within the Jaeggli and Safir system, forms with zero endings, such as the 3rd person singular of the Serbo-Croatian verbs above, pose a potential problem. Recall that licensing in this framework is the result of consistency in the entire paradigm: Null subjects are licensed in languages with either only derived or only поп-derived verb forms. If a zero within a paradigm of otherwise consistently derived verb forms is truly an affix itself, then the paradigm can still be uniform. However, if the zero represents no ending at all, then the paradigm is derivationally mixed, and null subjects should not be licensed. The question is whether or not the potentially

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access problematic 3rd person singular forms ruča and govori in the Serbo-Croatian paradigms are bare stems. It turns out that we can claim in keeping with Jakobson's (1948) Hone stem system”

that these forms do actually involve morphologically active, or "real" affixes. First, there is evidence that the Serbo- Croatian form ruča involves a consonantal zero ending. Assum- ing that the basic stem is "ručaj-", we can account for the

3rd person singular present tense form by arguing that the consonantal zero affix of the 3rd person causes truncation of the jot due to the unstable C+C environment, just as the m of the 1st person singular or t of the infinitive causes truncation of the jot:

(37) ručaj + ti -► ručati (infinitive) ruča j + m -► ručam (1st sg) (ruča (3rd sg ־*- ,ruča j + 0i+cow

The jot of the basic stem is evident in the imperative form ru£aj. It is possible that such imperative forms really have no endings at all. However, that would be no problem for the morphological uniformity criterion, since the imperative in general seems to behave independently from the null-subject parameter. Note that even in canonical overt-subject languages

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access such as English, there is no overt expression of the subject of an imperative.15

Within the same system, the forms govori and govoriš in Serbo-Croatian actually derive in the following way:

(38) govori + i govori (3rd sg) govori + iš *♦ govoriš (2nd sg)

Supporting evidence for this analysis, in opposition to assuming "govori + 01+com|" and "govori + š", comes from verbs like

5 Feldstein (1987) comes to a comparable conclusion in his analysis of Polish czyta (read 3 sg), while he posits a "real" zero desinence for the imperative czytaj (imperative, 2sg.). Working with Jakobson's (1948) one-stem system, he posits "czytaj" as the basic stem. It has been considered a problem for the one-stem system to explain why a zero-ending causes the jot to truncate in one form (3rd sg) and not the other (imperative). Feldstein proposes that Polish verbal desinences in the non-past, including zero endings, can be further segmented into smaller units representing the conjugation class, person and number. Based on phonological evidence (i.e. the occurrence of interword sandhi) he assumes that there is a word boundary between the zero unit and the person and number endings in the imperative. He further proposes special truncation rules to deal with desinences which contain zero elements. If a zero is accompanied by a non-zero unit within the same word boundary, then the zero remains without any effect on truncation. If there is only a single zero within the word-unit, the zero desinence behaves like a vowel. This is the case in the imperative, and consequently the jot is not truncated. If there are multiple zeroes present within the word unit, they have the truncation effect of a consonant. This is the case in forms like czyta. The behavior of the different "zeroes" is illustrated below (# signifies a word boundary). czytaj +0+0+0# : czyta (C+C; truncation of "jot") czytaj + 0 # 0 + 0# : czytaj (C+V; no truncation) It is obvious how this analysis could be extended to Serbo- Croatian as well.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access želeti which obviously belong to the same conjugation class as govoriti:

(39) želim želiš želi želimo želite žele

,im־*} The personal endings of this conjugation type are thus -iá, -i/ imo, ite, e}, and the basic stems for the two verbs above are "govori" and "žele" respectively. Following Jakob-

son's rules for combining basic stems with endings, the final vowel of the basic stem is truncated due to the unstable V+V environment when the vocalic personal endings are added. The infinitive ending {-ti}, on the other hand, is consonantal» and no truncation occurs when it is added to the vocalic basic stem.

I assume that the Czech form dëlà (3rd singular) also involves a zero consonantal affix. All the forms of the paradigm above can be derived from the basic stem "dëlaj", provided that there is a special rule of vowel lengthening in Czech in C+C environments in the non-past paradigm. This is illustrated in (40) .

(dëlàm (truncation; lengthening ־*- dëlaj + m (40) dëlaj + 0i+comI dëlà (truncation; lengthening) dëlaj + £ - dëlají

We have seen that there is nothing in the present tense paradigm of Russian which would account for the lack of identification of null subjects. However, there is a crucial

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access difference in the past tense formation between East Slavic on the one hand and West and South on the other, which will be discussed below.

Historically, Old Church Slavonic (OCS) was pro-drop, while modern Russian is not. The question is, then, what happened over time which could have caused the loss of pro- drop in Russian. The answer lies in the development of the past tense(s). OCS had four different past tenses: The aorist and the imperfect, which were simple tenses, and the perfect and the past perfect, which were compound tenses. Many Slavic languages today, including Russian, have only one past tense, which is derived from the perfect tense of OCS.16 The perfect tense in Slavic was originally formed analytically with a

present tense form of byti (1to be'), indicating the grammatical categories of person and number, and the short form of the !*participle which showed agreement in gender and number with the subject, as shown in Table 3.17

16 The decline of the aorist and imperfect took place relatively early in Russian. It is difficult to give an exact time frame, since this process is reflected differently in different manuscripts. Also, the bookish church style seems to have been much more conservative than the secular (delovoj) style, where these tenses become less frequent at an earlier point in time. Ivanov (1990) gives the 13th/14th centuries as an approximate time for the considerable decline of the imperfect and the aorist in Russian.

17 The dual number is not crucial for the discussion and is not listed in Table 3.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-60- Table 3: OCS Perfect Tense Sg. mase • fern. neuter

lat jesmb xodilb xodila xodilo 2nd jesi xodilb xodila xodilo 3rd jestb xodilb xodila xodili PI•

lat jesmrb xodili xodily xodila 2nd jeste xodili xodily xodila 3rd SOtb xodili xodily xodila 4

Russian lost the original gender distinctions in the plural and the nasal vowels (stotb -* south) relatively early (•12th/13th centuries). Otherwise, the system above can be found in early Old Russian texts. Sentences (41) and (42) are examples of an intact, per- feet tense in Russian Church Slavonic manuscripts:18

(41) a. ne dalb jeei kozblçte (Ostromir's Gospel 1056-7) not gave аих^ goat •(you) have not given a goat' b. ѵёть jako otb boga jesi priãelb (ibid) know that from God aux^ arrived ' (I) know that (you) have come from God1 (42) jako že oci eoutb ustavili i zapovSdali as emph. fathers aux^ established and commanded tako že stvori samrt (Izbornik Svjatoslava 1076) so emph. self 'as the fathers have established and commanded so do yourself 1

Examples are cited from Ivanov (1990: 332-337) and Avanesov and Ivanov (1982: 92-106).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access However» the manuscripts of the 11th and 12th centuries already exhibit a frequent use of the perfect without auxiliary, of which (43) is the oldest attested example:

(43) glëbb knjazb тёгіГь more (Tmutorokan Stone; 1068) (b׳prince Gleb measured the sea' (missing auxJų jest,

(44) ježe otbcb dajalb i rodi sbdajali (B-bark #9; 1100s) 'what father bequeathed and relatives bequeathed* (missing aux^ jestb aux^ sotb)

The following examples taken from 13th and 14th century manuscripts show the continuing loss of the auxiliary in the perfect tense.

(45) a. otbcb moi оитегГь (Laurentian chronicle; 1377) father my died 'my father died' b. gdë že ne xodìli ni dēdi naši (ibid) where emph. not went nor granddads our 'where our grandfathers didn't go* c. ѵъ se že lēto V. ratb ройаГь (ibid) in this emph. year V. war started 'in this year V. started the war' (46) ne гекГь mi jesifb (В-bark #3; 14th cent.) not said me Joseph 'Joseph didn't tell me' (47) Sto selb pokupili bojare (Novgorod chart. 1368-71) that land bought Boyars 'and that Boyars bought the land* (48) bratb dSjalb nasilie (Novgorod charter 1264-65) brother did violence *the brother committed violence'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Several scholars, including Ivanov (1990) and ternyx (1954), point out that this loss of the auxiliary was particularly frequent in the 3rd person with overt subjects. This is re- fleeted in examples (48-51) above, which all have full NPs as subjects. According to Cemyx, the Smolensk Charter from 1229 uses the perfect tense of the 3rd person 23 times, but not a single time with the copula. There is a good reason why the loss of the auxiliary seems to prefer the 3rd person. The exam- pies from the manuscripts show that the copula was dropped when the loss of the pronominal features was compensable by the occurrence of an overt subject. If we assume that Old Russian was still pro-drop at that stage, then overt subjects (pronouns) of the 1st and 2nd person were only used in emphatic contexts. For the 3rd person, on the other hand, any full NP could act as the поп-emphatic subject and thus the instances of sentences with overt subjects in the 3rd person were much more frequent. Although the auxiliary was preserved longer in sentences with 1st and 2nd person subjects, there are also early examples for the loss of the copula in these persons, as illustrated below.

(49) ja dalb rukoju svojeju (Mstislav's charter 1130) 1I gave with my hand' (missing aux!u jesmt)

(50) ty овкІаЬіГь sę (Antiouchus Pandects 11th century) 'you smiled' (missing aux^ jesi) (51) my vašee bratije ne obidëli ni grabili (Bishop Jakov's letter to Riga, 1300) 'we didn't offend or rob our brothers' (missing auxIp1 jesnth)

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

63-־ The original meaning of the perfect tense was that of an event which started in the past and which continues into the present. Formally this "bridging of time" was conveyed by the present tense of the auxiliary. According to Avanesov and th centuries־Ivanov (1982: 92-107), the documents of the 1214 reveal that this original meaning of the perfect tense was in transition at that time. In this context they cite several examples of formally intact perfect tenses with simple past tense meaning. The gradual loss of the auxiliary as described

above may have been both a contributing factor to and a logi-

cal consequence of this process. The other purpose of the auxiliary, namely that of expressing the category of person, is perhaps more important for our investigation. As mentioned

earlier, this role of the auxiliary was compensated for by the use of an overt subject, NP or pronoun.19 As the perfect tense became established without the auxiliary, overt pronouns became disassociated from their emphatic function and started to spread widely in Old Russian texts.20 Ivanov points out

Old Church Slavonic did not have a 3rd person per- sonai pronoun. 3rd-person reference was made by the demonstratives tî>, зь and опъ. The latter developed into the 3rd person pronoun.

20 The shift in the meaning of the perfect tense cannot be interpreted as the sole reason why East Slavic developed поп-emphatic personal pronouns and thus lost its pro-drop properties, since the same change happened in other Slavic languages which remained null-subject. It seems that all Slavic languages which use the 1-participle as the simple past tense form of the verb could have gone one way (keep the auxiliary) or the other (use поп-emphatic subjects pronouns), once the "bridging function" of the auxiliary was lost. In Polish, the auxiliary forms of the OCS perfect tense have Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access that the formally intact perfect is still occasionally found in later manuscripts (14th and 15th centuries), especially in the religious style. Interestingly, we can find such examples of the analytical perfect tense both with and without overt subject, which shows that the use of (поп-emphatic) overt pronouns (probably due to the loss of the auxiliary in normal speech) had already gained in popularity.

(52) sego ty že esi хоЪ&Гь (Laurentián Chronicle 1377) this you emph. aux^ wanted *you wanted this'

In summation, it seems plausible to attribute the loss of (referential) pro-drop to the reinterpretation of the 1-participle as a verbal form, which due to its participial origin, is somewhat peculiar in shape: Contrary to other forms of the verbal paradigm, it does not show person, but it does show gender instead.21 been incorporated into the verb form, functioning as desinences which represent the category of person/number, very much like desinences on present tense verb forms. pytać ('to ask1) present past lsg pytam pytałem 2sg pytasz pytałeś 21 The fate of the past perfect, the third "extinct" tense, was indirectly influenced by the development of the other past tenses. Originally, the past perfect was formed by the imperfect of the verb byti and the 1-participle (e.g. nesli bëste 'you-all had carried'). In the 13th century the past perfect started using the perfect tense of byti instead of the imperfect (e.g. nesli jeste byli 'you-all had carried'), as shown in (i). This is presumably connected to the loss of the imperfect tense at around the same time.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access In contrast to East Slavic, West and South Slavic did not drop the auxiliary verb in the formation of the past tense, so that the auxiliary has preserved its purpose of expressing the category of person/number and the personal pronouns their emphatic function, as was true for OCS. This is shown below for Czech and Serbo-Croatian.

(53) Cetla jeesa tu knihu [CZ] aux,u this book ,״,read 1I read this book' (54) Sitala 8am tu knjigu [SC]

3.2.2. How to foraalize the difference between Rueeian and West and South Slavic

It would seem plausible that the special development of the past tense would have necessitated the use of overt pronouns in that tense only in Russian. However, thematic null subjects

(i) a čto selb i svobodb ... 1and as far as their land and freedom ...' to dali jeeme byli andrëju that gavėju auxlp( auXj^ Andrej 'we had given it to Andrej1 (Novg. charter 1304/05) At about the same time, the perfect starts to lose its auxiliary, which is also reflected in the past perfect. This is illustrated in (ii). (ii) a ot петесъ byli na sude iekali kolokola and from Germans auxi1mt at court searchedj^ bell (1284; Gramota knjazja Fedora Rostislaviča) 'but they had demanded the bell from the Germans in court' The past perfect in this latter form is preserved until the 17th century, but is missing from contemporary literary Rus- sian.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access in general stopped being identified, even in the present tense where unambiguous identification never seems to have been at risk.22 This may imply that the whole paradigm, and not parts of the paradigm or even individual verb forms, determines whether empty subjects are possible or not.23 The 3rd person deserves further comment. In Slovene, Serbo-Croatian, and both Sorbian languages the copula is used in all forms of the compound past (perfect) while the other South and West Slavic languages have a zero copula in the 3rd person. This is illustrated below.

(55) ljudje •o videli [SLN] people аиХф sawp, Janez je videi Janez auxj^ s a w ^

(56) ljudi eu videli [SC] people аихзр! saw^, Jovan je vi deo Jovan auxj^ saw^

22 Recall that following Jaeggli and Safir (1989) we differentiate between "licensing" and "identification" of null subjects. All null subjects have to be licensed, but only thematic subjects have to be identified. 23 A hypothesis stating that this would be universally true is inconsistent with the pro-drop properties of Hebrew (among others). In Hebrew, null referential subjects occur only in the past and future tenses, presumably because the present tense AGR is not sufficiently rich to identify pro. The situation is further complicated by the fact that 3rd person referential pro is licensed only in embedded contexts with an antecedent for the pro-element in the matrix clause. Pro in matrix clauses is possible only in the 1st and 2nd persons. For a possible analysis see Borer (1989).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access (57) luźe au wiźeli [LS] people auxjp, sawp, Jan jo wiźeł ^״Jan aux^ saw

(58) lidé vidëli [CZ] people sawp! Jifi vidël Jifí s a w ^

(59) ludzie widzieli [PO] people saWp, Jerzy widział Jerzy saw«,

The verbal agreement feature of person is nevertheless also available in all South and West Slavic languages in the

3rd person: A zero copula signals 3rd person. This shows that richness is not necessarily identical to overt expression of agreement features on a given verb form. In other words, it is not the lexicalization of AGR on an individual verb form which makes an empty subject possible. It turns out that to formalize what exactly blocks identification in Russian is a fairly difficult task. Since Russian actually licenses expletive subjects, an analysis à la Speas (1994) is bound to fail. Recall the split INFL hypothesis, according to which IP is decomposed into two functional projections: AGRP and TP. Speas further argues that, according to the principle of economy, a projection XP can be projected only if it has content. Since AGR has no semantic content, AGRP can be projected under two conditions: The head itself is

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access filled by an agreement affix, or the specifier position is filled.24 Returning to the Russian problem, we could assume, adjusting the definition of "strong" agreement appropriately, that Russian is a language with "weak" agreement, due to the lack of agreement morphology in the past tense.25 The verbal affixes would then not have independent lexical entries, but instead be listed in the lexicon in the verbal paradigms. In keeping with the principle of economy, the specifier of AGRP would thus have to be filled, disallowing pro. However, given this hypothesis, we could not account for the occurrence of null expletives, unless we were to assume that this phenomenon

2 The 3rd person forms without copula in West Slavic might be problematic for Speas's analysis. According to Speas (1994: 8) the relevant notion of content is the following: A node X has content if and only if X dominates a distinct phonological matrix or a distinct semantic matrix. Forms like Czech Seti (*he read1) or Polish czytał do not have an overt person/number agreement affix (I assume that -ł is the head of TP) . Consequently, following the economy of projection principle, we would be forced to assume that AGRP is not projected at all when these forms are being used, since the AGR does not dominate a phonological matrix and there is no subject occupying the specifier position. However, this theory contra- diets the intuition that the agreement features are really present and available, though not phonologically present. 25 Under Speas's criteria (using Rohrbacher's (1995) "conditions for lexically listed affixes", cf. Chapter 2) Russian could also be analyzed as a "strong" AGR language, thus theoretically allowing null subjects. However, Speas also mentions explicitly that a "strong" AGR language may choose to list agreement affixes as part of the verbal paradigms, thus forcing pronominal subjects to be overt. While this also raises the question of how to account for null expletives, the main problem with Speas's analysis, from my perspective, is that it is not specific enough to say anything interesting about why Russian is different (See Chapter 4, fn. 35 for a discussion of Sorbian within Speas's system) .

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access in general were due to some factors unrelated to the issue of pro-drop.

In Franks (1990), Lindseth and Franks (1994), and Franks (1994) an analysis within the licensing and identification approach is proposed. Russian fulfills the morphological uniformity criterion and therefore licenses null subjects. The lack of identification in Russian can be formulated in terms of an anomaly in the functional projections of V. The analysis corresponds to the intuition that the acceptability of pro should indeed be formulated in terms of richness of inflection. We have seen that although Slavic verbs may ex- and gender-number ( ״hibit both person-number agreement (AGRp agreement (AGR^), this is realized differently in East Slavic, compared to West and. South Slavic. Whereas AGR^ is consistently marked in South and West Slavic, in East Slavic it is absent in certain forms. Specifically, there is no present tense copula, and hence, as we have seen above, all verbs in -agreement morphology. In East Slav ,״the past tense lack AGRp ic, the two agreement types are mutually exclusive, which, in principle, would allow us to associate either type with what is generally known as the AGR (or AGRS) projection in the

and AGR^ in West and South Slavic ״,literature. However, AGRp can be present at the same time (such as in the past tense) and occupy distinct positions. The former is a relatively

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-70- free clitic, while the latter appears unequivocally as an ending after the tense suffix -I.26 This constitutes evidence that in a full-blown functional projection system the two agreement types should be two distinct heads. The Slavic clause structure could then be illustrated by the following tree structure in (60), whereby the relative order of the different functional projections may vary from language to language.

(60) CP A c '' A G R ^ P

,״י S p e c " AGRp AGR^P ״AGRį

S p e c a 'G R ^ '

AGRp^j TP

щ. Spec T ' T VP

Only AGRp^P seems to be relevant for the identification

of referential null subjects. In West and South Slavic languages null subjects are identified despite the lack of AGR^

in all forms of the present tense. Thus, a correlation between the availability of a null referential pro subject and the presence of AGR^ is proposed. Under this analysis AGR^P is

26 Even in Polish, the personal endings of the Past tense can be transferred and attached to other words in the sentence : Gdzie byłaś? or Gdieâ była? (,where were you1)

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access associated with what is usually called AGRSP. The Russian present tense forms show full person agreement, and yet pro is no more appropriate in present tense sentences than it is in It seems to be the absence .,״the past tense, which lacks AGRp

in the past tense which is responsible for blocking ״,of AGRp the identification of referential null subjects in general in Russian. It should be stressed once more, as already indicated above, that what is important under this analysis is the role of AGRp^ in the inflectional system, rather than its presence in individual forms.

3.3* Summary

In this chapter I applied three null-subject diagnostics to the Slavic languages. By these criteria, Czech and Serbo-

Croatian, as representatives of West and South Slavic, are canonical null-subject languages, while Russian is not. The reason why Slavic languages divide along this line is found in the different development of the past tense. Specifically, East Slavic lost the present tense copula, which led to a loss of person agreement in the past tense. The loss of the auxiliary started in sentences with overt subjects (full NPs or emphatic 1st and 2nd person pronouns), but eventually spread much further and was compensated for by the use of overt pronouns in general, which thereby lost their emphatic character. As a result, the division of labor between overt and zero pronouns, which is characteristic for null-subject languages.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-72- was lost. This insight was formalized by proposing two dif-

ferent functional projections AGR^P and AGR^P for the differ-

ent types of agreement in Slavic. Only the former is true

subject agreement in the traditional sense (AGRS) and relevant

for identification of null subjects. The inconsistent realiza-

was found to be the culprit for the blocking of ״tion of AGRpj

identification in Russian.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 4. IS SORB IAN A Pro-DROP LANGUAGE?

The null-subject diagnostics previously applied to other Slavic languages produce the unexpected result that Sorbian is not a canonical null-subject language. Three possible expia- nations for this are explored in section 4.6. While all three are workable in principle, the last of these, based on a competition model of language change (Sprouse & Vance 1995), proved to present the fewest problems and give the most plausible account for the whole range of Sorbian data. This is partially due to the fact that it incorporates the broad historical context of the language.

4.1• The Sorbian population of Germany Soon after the reunification of Germany, in 1991, a political scientist from Bonn wrote:

Sorben sprechen fließend deutsch, sie besitzen deutsches Wahlrecht und deutsche Pässe, sie sind deutsche Staatsbürger -- aber sie sind keine Deutschen! Deutsche Kaiser, Führer und General- sekretāre haben versucht, ihnen die Luft abzu- drücken -- mal im Würgegriff, mal in der Umarmung. (Oschlies, 1991: 2)'

In general terms this 1991 commentary contains some basic points which characterize any study of the Sorbian population

Sorbs speak German fluently, have German voting״ 1 but they are ־- rights, German passports, are German citizens not Germans. German emperors, "Führer" and General Secretar- ies have tried to suffocate them -- sometimes in a stranglehold, sometimes in an embrace.1

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-74-

of Germany, and as will be evident below, this is also relevant for the field of linguistics.

The Sorbs today are a national minority in a very small piece of territory called die Lausitz (Lusatia) in what was

formerly the German Democratic Republic (East Germany). In

English they are known variously as Sorbs, Lusatians or Wends. In this work I will follow the current style and refer to the

language as Sorbian, the people as Sorbs, and the territory as Lusatia. The Sorbs refer to themselves as Serbja (Serby in Lower Sorbian) and are the last survivors of the Slavonic tribes which once occupied most of the territory between the Oder and Elbe rivers. They are Slavs, but unlike other Slavic nations they have no separate political existence and are not a sub- group of any larger ethnic identity outside Germany. They as distinct as the ־־ want to be thought of as a Slavic people -minor" only in a quantita״ -- Russians, Bulgarians, or Poles tive rather than qualitative sense. However, even in Germany there are many people who have never heard about the Sorbs. In fact. Western tourists in eastern Germany were known to panic when they entered the bilingual Lusatian region, believing that the Sorbian language place-name signs meant they had unwittingly crossed the border into Poland. Moreover, even scholars who have specialized in the study of the Czech and Polish languages admit to knowing little about the Sorbs, some having been taught that Sorbian was a dead language.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access The area where one can still hear Sorbian spoken currently begins about 50 miles south-east of Berlin and stretches approximately 55 miles to the south. It is about 40 miles across at its widest point with Cottbus being the center of Lower Lusatia (so named because it is downstream on the River Spree) while Bautzen is the principal focus of Upper Lusatia. The modern regions of Upper and Lower Lusatia correspond roughly with the areas occupied by the Milzane and Lusizi peoples, respectively, prior to German immigration during the 10th century, and the historical delineation of two Sorbian literary languages appears even in the earliest translations of religious works during the 16th and 17th centuries. However, it was German political development in the 17th century which finally separated these regions and encouraged the emergence of two major cultural*and economic centers: Cottbus in Branden- burg/Prussia and Bautzen in Saxony. As a result, the dialects spoken in these two places were codified as the modern literary languages. It is difficult if not nearly impossible, to find reliable information about the number of Sorbian native speakers. The Communist government of East Germany was supposedly not interested in research which could establish such data, since the results might denigrate the officially proclaimed "sozialistische Nationalitätenpolitik” by revealing the pro- gressive assimilation of the Sorbs. According to a 1987 projection by the Institut fúr sorbische Volksforschung, the

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access -76- number of people who knew Sorbian was estimated at about 67,000 (16,200 of which were speakers of Lower Sorbian). However, this figure included a large number of persons with only very limited language skills, individuals who learned Sorbian in school as a foreign language, and some who did not use it at all as a language for everyday communication. Re- cent study indicates that it is only in the comparatively tiny Catholic area of rural Upper Lusatia (to the north-west of Bautzen and roughly defined by the villages of Radibor, Crost- witz, Panschwitz-Kuckau, Ralbitz and Räckelwitz) that one can still find predominantly Sorbian-speaking communities. Ac- cording to Elle (1991), the 15,000 Sorbian speakers of this Kerngebiet" (core area) make up over % of the total״ Sorbian number of people who actively use Sorbian today and thus could be considered native speakers. It can then be deduced that the total number of native speakere of Sorbian must be around 22,500. According to Spieß (1995) there are about 5,100 native speakers of Lower Sorbian, which leaves about 2,4 00 native speakers of Upper Sorbian outside the Catholic "Kerngebiet". Elle (1992: 73-79) points out that in the "Kerngebiet" 96% of the people under 3 5 who know Sorbian learned it at home, 95% speak Sorbian in their own families, and 76% speak only Sorbian (never German) at home. In the other (much larger) region of Lusatia only about 29% of the people under 35 who know Sorbian learned it at home, and only 9% use it today in their families. This is somewhat surprising, considering that

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access over 90% of the ethnic Sorbs in that region have at least some degree of knowledge of Sorbian. Also, there is a significant difference between generations, which is not the case for the "Kerngebiet " : 49% of the people over 59 who know Sorbian still use the language at home, while for the group between 36 and 59 this figure falls to about 22%.

The numbers given above show clearly that most regions of Lusatia have succumbed to German assimilatory pressure and are

German dominated mixed-language territories at best. This mixture is hardly surprising since German cultural pressure has been a constant force since the 10th century, though at that time Lusatia still had substantial contact with her Pola- bian neighbors to the north, the Bohemians to the South and Poles to the East. Continued German expansion, however, meant that by the 15th century the Sorbian speech area was surrounded almost entirely by German speakers. Despite almost 1000 years of Germanization, however, Sorbian is by no means 1*dead". The

Sorbian publisher Ludowe nakładnistwo Domowina still publishes a daily Upper Sorbian and weekly Lower Sorbian newspaper, as well as the academic journal Lëtopis and the pedagogical journal Serbska Šulą. There is also a significant number of contemporary popular and children's books published every year, and the academic and reference works by researchers of the Sorbische Institut in Bautzen and the Institut für Sorabistik at Leipzig University are among the most complete for any Slavic language. There are also Sorbian kindergartens and

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-78- schools, and a Sorbian-language radio station affiliated with Mitteldeutsche Rundfunk.

4.2. Verbal Morphology of Upper Sorbian (Upper) Sorbian is among the more archaic of the modern Slavic languages. For instance, it has preserved the use of the Dual number, and the Aorist and Imperfect tenses from Common Slavic.

The present, perfective future, aorist, and imperfect are expressed synthetically. The aorist is formed from perfective verbs and the imperfect from imperfective verbs. The endings are the same for both tenses with the exception of the 2nd and 3rd person singular, and several grammars treat both tenses together as the simple past tense. The perfect, past perfect, and imperfective future are expressed analytically. The per- feet tense is formed with the present tense of the verb być 1to be1 and the I-part iciple, which is analogous to Old Church Slavonic (OCS). Similarly, the past perfect tense takes the imperfect of być and the I-participle.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־79־ Т а Ы е 1: Verbal Paradigm of Upper Sorbian džētač = ,to work1 Present/Future Imperfect Aorist Sg. Ī"! (na) džēłam džēTāāh nadžēlach 2. (na)džēiaš džēlaše nadžēia 3. (na)džēla džēlaše nadžēla Pl. ī~. TnaTdžlTimy džēlachmy nadžėiachmy 2. (na)džēlače džēlašče nadžēiašče 3. (na)džēlaja džēłachu nadžēlachu Dual na) džēłamoj džēTāāhmoj nadžētachmoj) .־ T atej2 džēlaštaj/-aštej nadžēlaštaj/-aštej־/na) džē łataj) .2 aštej nadžēiaštaj/-aštej־/na) džē łataj/-atej džēlaštaj) . 3 1-participle: mase. fern. neuter

Sg. (nāTdžēTāT (na) dźSłała (na) džē łato Pl. (na)džētali Dual (na)dźółałoj b y ć = ' to be1 Present Imperfect Sg. 1. sym bëch 2. sy bē (Še) 3. je bë(äe) PI. 1. smy bëchmy 2. sće bēŠče 3. su bëchu Dual 1. smój bëchmoj štej־•/staj/stej3 bēštaj .2 3. staj/stej bēštaj/■-štej

Virile vs. non-virile distinction.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Kxamplaa of Perfect:

ja sym džēlata I have been working1 mój smój džēlaloj we two have been working Examples of Past Perfect: י ja bëch džētala I had been working mój bëchmoj dź$łałoj we two had been working

Building on the discussion in Chapters 2 and 3 we can see that the Upper Sorbian verbal paradigm meets the morphological uniformity condition, since all forms are consistently derived. Thus, Sorbian should license null subjects. It would also seem that the "richness" of the Sorbian verbal agreement system can easily compare with that of the other West (and South) Slavic languages. In Chapter 3 we established that Czech (as

a general representative of West Slavic) qualifies as a canonical null-subject language. Consequently, we would expect the same to be true for Upper Sorbian. Also, since Common Slavic was a pro-drop language, and the Sorbian verbal system seems to preserve the Common Slavic situation as well as any of the other Slavic languages, the expectation that Sorbian is a null-subject language is reinforced.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 4*3• Grammar v e • Reality

Upper Sorbian has personal pronouns for the singular, plural, and dual.

Table 2: Personal Pronouns of Upper Sorbian 1. pers. 2. pers. Sg. Jā tÿ Pl. my wy Dual mój wój

3 . pere. mase. fem. neuter Sg. wón wona wono/wone PI. woni, wone Dual wonaj, wonej

The dialectal distribution between the two neuter forms

wono and wone will be discussed in Chapter 5. The plural form woni is used to refer to males, whereby one individual in the group is sufficient to use this form of the 3rd person plural pronoun. The form wone is used elsewhere, and can refer to people, animals and inanimate things. The same distinction is expressed in the dual. According to current school grammars and older influential grammars such as Jordan (1841), Schneider (1853), Liebsch (1884) and Krai (1895), personal pronouns in Upper Sorbian should only be used in emphatic contexts. Šewc (1984: 117), for instance, writes:

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Na rozdžēl wot nëmskeje a ruskeje rēče wužiwa so w serbščinje forma personalneho pronomena pfi werbach jako subjekt jenož we wobmjezowanym rozsahu. Zwisuje to z tym, zo charakterizuja same wosobowe kóncowki hižo jasnje dość wotpowëdnu wosobu.

,In contrast to German and Russian, the personal pronouns in Sorbian are more restricted in their use. This is connected to the fact that the personal endings themselves already indicate the corresponding person quite clearly.1

This is certainly reminiscent of the description of personal pronoun use in null-subject languages such as Czech or Serbo- Croatian discussed previously.

In most of the older grammars the authors are even more outspoken about the lack of overt pronoun use in Upper Sorbian. For example, Jordan (1841: 169) writes:

Da die Personen an dem Ausgange des Verbums hingâng- lich erkennbar sind, so werden die persönlichen Für- Wörter ja, ty, wón, mój, wój , wonaj , my, wy, woni nur dann gesetzt, wenn die Person besonders ausge- zeichnet werden soll.4

Similarly, Schneider (1853: 250) writes:

... die persönlichen Fürwörter: 'ich, du, er, wir, ihr, sie' werden im Wendischen nur dann gesetzt, wenn ein besonderer Nachdruck darauf ruht, oder wenn ein Gegensatz angezeigt werden soll. Z.B. Ja sem to pisaw, nic té ich habe das geschrieben, nicht du.5

*Since the person is sufficiently recognizable from the verb ending, the personal pronouns j a , ty, wón, mój, wój, wonaj, my, wy, woni are only used if the person is to be especially marked.*

5 the personal pronouns: 'I, you, he, we, you, they1 are used in Sorbian only if they carry a special emphasis or if they indicate a contrast. For example. Ja зет to

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Liebsch (1884: 27/28) gives the same rule even more explicitly:

Dieses Verhältnis . . . wird im Wendischen ähnlich wie im Lateinischen und Griechischen durch die Personalendungen des Verbum bezeichnet. Deshalb ist die nochmalige Bezeichnung desselben durch die Personalpronomina durchaus überflüssig und nur dann notwendig und gestattet, wenn wir das Subjekt aus irgend einem besonderen Grund nachdruckevoller hervorheben wollen . .. Das ausdrückliche Bezeichnen der Person durch das persönliche Fürwort im Wendischen ist daher nur ein Germanismus.6

Finally, Kral (1895: 58) writes:

Die persönlichen Fürwörter als Subjekte werden nur dann gesetzt, wenn auf denselben ein gewisser Nachdruck liegt, oder wenn sie im Gegensätze stehen.7

Thus our expectation that Upper Sorbian is a canonical null-subject language seems to be met. This is confirmed by the following example from contemporary literary Upper Sorbian. The excerpt is taken from Jurij Koch's novel Ha lećała je módra wróna published in 1991. (# indicates all places where a subject is "missing")

pisaw, nic ty ('I wrote that and not you').1 6 1This relationship ... is expressed in Sorbian, very much as in Latin and Greek, by the personal endings of the verb. Therefore, the additional expression by personal pronouns is completely superfluous, and only necessary and permitted if the subject is supposed to be stressed more emphatically. . . . The explicit marking of the person by a personal pronoun in Sorbian is thus just a Germanism.1 7 'Personal pronouns as subjects are only used if they carry a certain emphasis, or if the are used to contrast.'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access -84- Moja maćerna rēč je serbščina ... Z tutej rēču sym # wotrostł. # njemóźu so wjace wot njeje džēlič, tež nic, hdyž bych # to chcył. # Njejsym sej ju wupytał. # Bë tu hižo, hdyž # dóńdźech. 1My mother tongue is Sorbian ... With this language I grew up. I cannot part from it any more, not even, if I wanted to. I did not select it. It was here, when I arrived.1

However, Upper Sorbian dialects provide us with contradictory data. Published transcripts of dialect texts reveal a surprising level of overt subject pronoun use, despite the fact that the dialects have the same rich person/number system as the standard language. The following are some typical examples. Sentences (1-3) are taken from Faßke and Michałк (1989), while (4) and (5) are from Michałk and Protze (1974) .8

(1) a do pównoce smó mó khodźili and till midnight auxlpļ we went9 *we walked until midnight' (2) tuijs sej woni Žane (j) rade wejdźili nejsu also refi they none advice knew neg-aux^ ,and since they didn't know what else to do' (3) ale ja sn пёк tam wele moli ribach bów but I auxhg then there many times mushrooms was a ja (ja)c jich ničo widźiw nejsym and I more they^ nothing saw not-auxItg 'but I often used to go there for mushrooms and I didn't see any of them anymore'

* Some of these examples contain dialect forms which are not found in standard Sorbian. 9 The notation "auxlf|" etc. assumes present tense of the auxiliary być unless otherwise noted.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access (4) woni praja, woni dyrbja tom knjez Šminderej, zo they say they must that^ Mr Šminder^ that woni, hejzo jich njepušča, woni dom njepóńdźeja, they if them not-admit they home not-will*go d6jž j oh woni njeńdźa widźeć go see־will־until him they not ,they say they must see Mr. Schminder, that they, if they are not admitted, they won’t go home, until they have seen him'

(5) hale kóźdy mól pak bóch je cedlku pfitykwa; but every time but аихІ1|ИІЧ I note included; ja sym jare rjenje pisawa very nicely wrote ״I auxIł 'but every time I included a small note; I wrote very nicely'

An interesting descriptive discussion of the problem is provided by Faska (1964). He concludes from an analysis of a large number of recordings of speakers from various regions that the traditional "null-pronoun rule" conflicts with actual use in Sorbian. In fact, in the sample he analyzed, there were more instances of overt subject pronouns in tensed clauses than instances of zero pronouns. It is not surprising that overt pronouns are used in emphatic and contrastive contexts in spoken Sorbian. However, Faska (1964) also determines other environments where spoken Sorbian more-or-less consistently uses an overt (поп-emphatic) pronoun, and it is these instances, illustrated in the examples (6-8), which are of particular interest to us.10

10 All examples cited here are from Faska (1964); some contain dialect forms which are not found in standard Sorbian.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-86- (6 ) If direct or indirect speech is being introduced: a. "ja so ničo njeboju", je w6n rjekł 'I don't fear anything1, he said b. a ja njebëch tak hłupy, ja sym pïeco prajił, zo ja dale na dźSło póńdu do mèsta 1but I was not so stupid, I usually said, that I would keep on going to town to work' c. woni su mi prajili, št6 je to był, ale ja sym to zabył 'they asked me what that was but I forgot' (7) If the matrix subject is a demonstrative pronoun: a. t6n te dwaj abo tïi dny spać njem6Žeše, dokelž wón dyrbješe wšo widźeć a wjeść 'that one couldn't sleep for two or three days, because he had to see and lead everything'

b. a ta je so tak wustróźała, zo je wona chroma wostała 1and that one got so startled that she stayed lame1

ז If the subject of the matrix clause is л full N P (8)

a mój nan je prajił, zo bdźe wón pfećiwo Vogtej hić 'and my father said that he would go against the overseer'

In most other environments (except for imperatives, impersonal sentences, and sentences with 3rd person arbitrary reading11 which strongly prefer or even require the subject to be zero) the pronoun is sometimes used and sometimes not. No general pattern is readily apparent, and the choice seems to

See also section 4.3. for the possible use of overt woni in arbitrary plural constructions.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־87־

depend on a variety of factors, (including the speaker and the rhythm of the sentence) . According to Faßke, of the 190 exam- pies he examined which were not instances of one of the environments above, 75 had a zero subject, while 115 had an overt subject. He thus concludes that it is not necessary to pro- hibit the use of overt pronouns in the literary language cate- gorically ("njetrëbne je, wužiwanje personalnych pronomenow w literarnej rēči tak kategorisce zakazać"). Jenč (1959) is mainly concerned with the position of the

auxiliary and the 1-participle in Sorbian. However, he also

provides some interesting data about the use of personal pro-

nouns in subject position. His investigation is based on a

large sample of native speaker recordings. The most important results with respect to our topic are the following. The forms of być are quite consistently treated as enclitics, and colloquial speech (ludowa rēč) does not put the auxiliary in initial position except in yes/no questions. The auxiliary is typically preceded by a personal pronoun, as the following examples, taken from Jenč (1959), illustrate:

(9) ja sym tam wjele moli był I auxllg there many times was II have been there many times'

(10) wón je sej tola myslił, zo wón te he aux3tf refi though thought that he this šerjene wjedźe ghost leads 1 he however thought that he leads the ghost1

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־88־ The number of instances where the auxiliary is found in initial position is low. According to Jenč this is mainly the case when people try to make a conscious effort to speak "correctly", and therefore without pronominal subjects. Example (11) below reveals this "extra" effort since virtually all Sorbs normally use German numbers in spontaneous speech.

(11) sym rodźena w Jaseńcy sydomadwaceteho septembra dźewjatnaće stow a wósomasydomdźesat ' (I) was born in Jeßnitz on September 27, 1978 112

One individual seemed to use the construction with the auxil-

iary in initial position intentionally as a stylistic device of terseness, perhaps comparable to the "diary-style" mentioned in Chapter 2. If something other than the subject is in initial position, the subject pronoun is still usually lexicalized, following the clitic verb form, as shown in (12-14) .

(12) tarn sym ja bydlił there auxļsg I lived 'I lived there1 (13) a nëtk etej wonej wječor Słój and sometime aux?ibftł they^ evening went *and sometimes they both went in the evening1

(14) za sport sym ja wëzo pfeco zaj im mël for sports auxUf I indeed always interest had 1for sports I surely have always had an interest1

12 This must be a mistake, since the recordings were made before 1959. The speaker probably meant to say 1878.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Jenč points out that the few instances without personal pronouns are produced by people who have jobs in the public sec- tor and are more likely to try to conform to the school style in which they were trained. Even those individuals usually start using personal pronouns more frequently further into the recording when their style gets more lively and they become less driven to try to produce "correct" Sorbian.

With respect to subordinate clauses Jenč points out that there, too, personal pronouns are used more frequently than they are omitted. In his sample the ratio came out 64% : 36% in favor of lexicalized personal pronouns. Below are two of the many examples cited by Jenč:

(15) hdyž sym ja najprjedy stužit, da sym ja when auxls| I first served however auxItł I kolosai lochki był very light was *however, I was very light when I first served' (16) hdyž je wó n wumrël ,t o ja wëm when aux^j he died that I know 'I know when he died*

It should be emphasized once more that in Sorbian the use of overt pronouns is not obligatory, in contrast to German. Even though the frequent use of overt pronouns in Sorbian is usually blamed on German influence, there are some examples

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־90־ where Sorbian speakers omit pronouns in German, which further supports the claim that pronouns in Sorbian are optional.13 There might even be evidence within the literary language itself that the very limited use of overt pronouns is an artificial phenomenon, induced by prescriptive grammars and not in line with the real language. It is well known that clitics in some languages occupy the second position in the clause, the so-called Wackernagel position.14 According to Jenč (1959) and Michałk (1970) the present, past and conditional parts of the auxiliary verb być belong to the class of clitics and as such should not be found in the first position of a clause, assuming that Sorbian is a Wackemagel language.*5 As we have seen above in the discussion of Jenč (1959), that expectation is

Michałk (1977) cites the following two examples for missing personal pronouns in German under the influences of Sorbian: (i) # bin mit in die LPG eingetreten (missing Jch) '(I) have also joined the cooperative* (ii) Kuhstall ist noch nicht so, wie # sein muß (missing er) 'cowbarn is not yet such as (it) must be1 14 If there is more than one clitic, there are rules for the positioning of those clitics relative to one another. In Sorbian, according to Michałk (1965b), the relative order- ing is the following: part icle/by<5/dat ive/accusative/nominative haČ drje sym či je ja tež prawje wumył if part. auX|Sg you,* them** I also correctly washed 'if I indeed washed them the right way for you' 15 The status of the forms of być as enclitics is not absolute. In interrogative clauses the auxiliary is found regularly in first position both in colloquial speech and literary style, where it might be analyzed as proclitic.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access borne out in normal spoken Sorbian. Typically they are in second position preceded either by a nominal subject NP or by the personal pronoun. However, in the literary language there are many examples with an initial auxiliary.16 Jenč attributes this phenomenon to the fact that writers feel obliged to comply with the rule of the official grammar that personal pronouns in general must not be used. Example (17) with the conditional form of być is taken from Koch (1991) , and (18) and (19) are taken from the newspaper Nowa Doba as cited by Jenč (1959) .

(17) bych sej samomu zakladne dožiwjenje aux1**wbj refi,* seifen basic experience pokradył; bych je pódla sebje stejo wostajil deprived; auxljgt^ it next myself standing left^ 1(I) would deprive myself of my basic experience; (I) would just turn my back on it1 (18) symprajił, zo doma wostanjetaj ,״^aux)l|pfa said that at-home will-stay ' (I) said that the two of them will stay at home' (19) bS nam te2 najnowêe nowiny sobu pfinjesł auxjggm^jcrfUS too newest paper self brought '(he) had also brought the newest papers with him1

Interestingly, even the puristic grammars of the 19th century still describe sentences such as the ones above as ungrammatical. Liebsch (1884: 228), for instance, writes that sym, ay,

This may also be the reason why some linguists, including De Bray (1963), write that auxiliary verbs in Sorbian are not clitics, in contrast to languages like Czech or Serbo-Croatian.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-92- je, smój, staj, stej t smy, sće, su must never be in the initial position in a sentence.

4.4. Application of null-subject diagnostics

Thus far we have seen that there is a contradiction between the normal spoken language and the exhortation found in vari-

ous Sorbian grammar books to omit overt subjects. From this

it is clear that we should at least be very suspicious about the pro-drop status of Sorbian. We will now apply the set of three diagnostics introduced in Chapter 2 in order to independently determine the status of Upper Sorbian with respect to null-subject characteristics. For the reasons discussed in Chapter 3 for Russian, this investigation will focus on the interpretive possibilities associated with overt subjects. The three diagnostics are repeated here as (20).

(20) The following holds true for null-subject languages: a. Overt subjects are stylistically marked. b. Overt subjects cannot function as bound variables. c. Overt subjects cannot be arbitrary in reference.

The linguistic data presented in this chapter come from various sources, which should improve reliability. I use examples based on published transcripts of spontaneous speech, examples cited in published articles, Sorbian sentences which I elicited from native speakers on the basis of German source texts, and Sorbian examples constructed by me and judged by native speakers. I conducted most of my fieldwork in the

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access rural area north-west of Bautzen which was previously referred Since there are only an estimated .*״ to as the "Kerngebiet 2,600 native speakers of Upper Sorbian dispersed outside this area, it was impracticable to undertake an exhaustive search. With respect to the null-subject criterion in (20a), the rampant use of unemphatic pronouns illustrated in the previous section strongly suggests that Upper Sorbian is not a canonical null-subject language. My fieldwork showed the same high level of overt pronoun use.

A similar conclusion can be drawn from the appearance of the overt wón 'he', wona *she* and woni 1they1 in the bound i). Example (21) is taken from a־variable contexts in (21a

(g־literary text and cited by Faßke (1981). Examples (21b were constructed by some informants when they were given the corresponding German sentences. All other informants accepted (21b-g) with a coreferential or bound reading of the pronoun in the subordinate clause when the sentences were given to them in Sorbian. Sentence (21h) was previously listed under (8) to illustrate one of Faßke's (1964) overt pronoun rules. In terms of our diagnostics, Faßke's rule means that a personal pronoun is usually realized phonologically if the subject antecedent of the matrix clause is a full NP, thereby directly contradicting the null-subject criterion (20b). Example (21i) is a sentence which Jenč (1959) uses in the explanation part of his article. Technically speaking, only (21b-d) are true examples of bound variables. However, criterion (20b) also

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-94- applies to structures where the QP is replaced by a referential NP (cf. section 3.1.2).

(21) a. Pelike njepytny, zo wón hižo hodźinu Feliks not-realized that he already hours po mēsče honja through town ran 1F. didn't realize, that he had already been running through town for an hour1 b. kóidy nan myśli, zo wón mudreho syna ma every father thinks that he smart son has 1every father thinks he has a smart son'

c. kóidy prošer so nadźija, zo wón skoro zbohatnje every beggar refi, hopes that he soon become-rich 1every beggar hopes that he will soon be rich1 d. kôïdy student sej myśli, zo wón jedynku dóstanje every student refi thinks that he A will-get 'every student thinks that he will get an A 1

e. Serb ja su hordźi, zo woni swoju rēč maja Sorbs auxjp! proud that they their language have 'Sorbs are proud that they have their own language' f. Jana je prajiła, zo chce wona rady kupać hić Jana aux3tfsaid that wants she gladly swimming go 'Jana said that she would gladly go swimming' g. Jana lubuje toho muža, na kotrehoŽ je so wona Jana loves that man at whom aux31g refi she pfed dźesać lëtami wudała ago ten years married 1Jana loves the man whom she married ten years ago' h. a mój nan je prajił, zo bdźe wón pfećiwo and my father aux3ų said that will he against Vogtej hić overseer go ,and my father said that he would go against the overseer '

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access i. Ale tež Grota staj i, hdyž w6n žiwišo rēči, but too Grofa puts when he lively speaks personalny pronomen, personal pronoun 1but also Grofa uses, when he speaks lively, a personal pronoun1

The application of the third null-subject criterion produces the same results, although perhaps less obviously so. While I did not observe the use of overt woni in spontaneous speech, there were some informants who used this construction when translating German sentences with arbitrary reference into Sorbian. However, the main point is that most inf or- mants, especially those without literary training, agreed that the presence of the overt 3rd person plural pronoun does not force a referential reading of the pronoun, in contrast to "true11 null-subject languages such as Czech. Examples of overt pronouns with arbitrary reference are given below.

(22) a. w Londonje su woni zaso stawkowali in London auxjp, they again struck 1 in London they went on strike again1 b. we Lužicy woni maja rjane stare naložki in Lusatia they have nice old customs 'in Lusatia they have nice old customs1 c. nëtko woni wšudže wo pfichodźe Serbów now they everywhere on future Sorbs^, diskutuja discuss ,nowadays they are talking everywhere about the future of the Sorbs'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access d. nžhdy su woni skromniâo 2iwi byli hač džensa once aux^ they modestly living were than today *once they used to live more modestly than today1 e. na tutej kupje woni hižo 500 lët francosce on this island they already 500 years French rëâa speak 1they have been speaking French on this island for 500 years*

f. na wikach woni sad pfedawaja at markets they fruit sell *they are selling fruit at the market* g. w Sewjernej Americe woni jendźelsce rēča in North America they English speak 'they speak English in North America'

I therefore conclude that Upper Sorbian is not a canonical null-subject language. This raises two main questions:

A. Why do the authors of most grammars try to describe Sorbian as a mill-subject Іавдалд• (In contrast to the actual epo- ken language) by strongly discouraging» if not even prohib- it ing, the use of overt non-emphatic pronouns? B. Why is Sorbian different from the other West Slavic (and South Slavic) languages?

The next section is devoted to the first question, while section 4.6. will focus on question B.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

97-־

4*5• Pronouns and puriem I agree with Faska (1964), who assumes that the pronoun-pro- hibiting rules of the traditional grammars were actually not formulated on the basis of the Sorbian language, but instead under the influence of Czech. Since the literary norm is based on the linguistic works of the middle and late 19th century, it is obvious how the pronoun prohibition rule made its way into current school grammars as well. This seemingly small detail of how to use overt pronouns can be seen as reflecting a much larger phenomenon when viewed in historical context. In the mid-19th century there was a strong Sorbian national movement known as the serbsJce narodne wozrodźenje (,Sorbian National Rebirth'), which led to a wide application of the literary language to secular situations. The Sorbs understood themselves to be part of the larger Slavic family of peoples, in keeping with the "Panslavism" of this period, and looked especially to their Czech neighbors as a guide. This was accompanied by the phenomenon of "purism",

aimed at complete removal of all foreign, and especially German* elements from the Sorbian literary language. New lexical needs were satisfied by the introduction of loan-words and loan-translations from other Slavic languages, especially Czech. Similar attempts were aimed at a removal of certain "non-Sorbian" morphological and syntactic phenomena from the literary language, most of which were attributed to German influence.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access alien" elements, which can be used״ One example of these to date the rise of purism fairly precisely (within 10 years), was the use of the demonstrative pronouns tón, ta, te in the same distribution as the definite article in German. As Lötzsch (1968) points out, analyzing different descriptions of the use of the definite article in Sorbian, Seiler's (1830) grammar is still free from puristic biases. Seiler tries to give an objective analysis of the use of the demonstrative pronoun as the definite article in Sorbian. Jordan's (1841) grammar, on the other hand, may be the first puristic work of Sorbian linguistics. He refers to the use of the definite -grundlose und wider״ and ״article in Sorbian as "Abnormalitàt sinnige Nachahmung des deutschen Idioms, die dem Charakter und dem Geiste der slawischen Sprachen rein zuwider läuft" (Jordan 1841: 138). Schneider (1858: 8) calls the definite article in ״.Sorbian a "Danaergeschenk der Deutschen17 Similarly, the use of overt pronouns in Sorbian has often Ibeen blamed on German influence, although this is not necessarily the case. We will return to the question of why Sorbian

1 The use of demonstratives as definite articles is still discouraged for the literary language but widely spread in spoken Sorbian. The following sentences are examples from colloquial Sorbian, taken from Michałк and Protze (1974). (i) ja sym słyśał, zo jo wón sej tu nohu złamał ich habe gehört, daß er sich das Bein gebrochen hat 'I heard he broke his leg' (ii) te ludžo su tola wáitcy wonka na namēsče die Leute sind alle draußen auf dem Anger 'the people are all outside on the village green־,

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access is not a pro-drop language in more detail below. Even if the use of overt pronouns proves to be genuinely Sorbian, language purism would have influenced this phenomenon because it appeared to be of German origin, simply because it was different from the pro-drop Czech (and Polish) neighbors. This turning to other Slavic languages to cleanse their own was a characteristic trait of the Sorbian purists. Language purism in general is very prescriptive in char* acter. and it contributed to a gap between the literary Ian- диаде and spoken language in Sorbian. This gap has only widened, since this crusade did not stop in the past century, unlike other puristic movements. Jenč (1959: 43) refers to this when he accuses contemporary writers of clinging to grammatical norms of other Slavic languages as if "hypnozowani" ('hypno- tized'). According to Jenč (1964: 1965), the difference between written and spoken Sorbian today is such that one might rather speak of two different languages instead of two different forms of the same language. He points out that there are only a few differences between the colloquial language today and the written language before the middle of the 19th century. We can thus assume that formerly there was no difference between spoken and written language and that it was (and still and "pure" Sorbian which has ״is) the search for an “ideal been moving them further apart by prescribing archaic and artificial norms for the written language. In several articles Jenč appeals to the Sorbian writers of today's literature

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-100- to overcome this schism and stop their Hwë6ne bojazliwe hladanje na druhe rēče" (1constant fearful look at other languages'). As Luther would have said, "man muß dem gemeinen Mann aufs Maul schauen" ('one should look at the common man's mouth') to

codify a language so that it can function as a widespread means of communication. The results of purism with respect to the use of personal pronouns in the written language can be illustrated by the following Biblical passage as it appeared in 1728 and 1896 versions, as cited by Jenč (1964: 252).

(23) a. Matthew 25:42,43 (1728, Richter (ed.)) .njejsće mi jësc dali ־Pfetož ja sym hłódny był a wy Mi je so pić chcyło a wy njejsće mi pić dali. Ja sym hósć był a wy mje njejsće hospodowali; ja sym nahi był , a wy mje njejsće wodźówali; ja sym chory a jaty był, a wy njejsće ke mni pohladali. b. Matthew 25:42,43 (1896, Lusćanski and Hórnik) Pfetož buch hłódny, a njedašce mi jësé; buch laćny a njenapojiŠce mje. Bëch cuzy, a njehospodowaśće mje; nahi a njewoblekaśće mje; khory a w jastwje a njewopytaSće mje.

There are nine overt personal pronouns in the version of this passage translated in 1728, none of which are present in Lusčanski and Hórnik's 1896 version in (23b). We have to assume that this change happened under puristic pressure since there are no indications that Sorbian had really become pro- drop (again).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Michałк (1972: 91,92) gives some interesting numbers with respect to the sudden decline of the use of overt pronouns in successive versions of Bible translations. The various Catho- lie translations seem particularly relevant since these are all based on the Latin Vulgate, i.e. direct influence from

German can be excluded. Swëtlik (Регііеорел18 1690) uses 20 personal pronouns in Matthew 2:2-18 which are not present in the Latin. In the 1750 pericopes, which differ only slightly from Swëtlik1s version, there are still 17 such pronouns. In the pericopes from 1848 this number is reduced to 7, in Buk*s (1862) translation there are only 3 left, and they are all stricken from Lusčanski and Hórnik's (1896) version.19 In the 1896 version we find that there are personal pronouns only in those places where they correspond to personal pronouns (1st and 2nd person), demonstratives or gui in the Latin version. Below are some examples:

Pericopes are selections of various Bible passages. 19 The Protestants translated their version of the Bible from Luther's German text. Interestingly enough, the number of поп-emphatic personal pronouns is actually lower than in the Swëtlik translation. In Frencel (1670) there are 13 personal pronouns in Matthew 2:2-18 which have no corre- spondence in the Homik/Lusčanski version. This number stays constant in the 1728 and 1823 versions, though in slightly different distribution. Obviously, the Protestants were not purifying" their Bible, because the 1905״ as concerned about translation still has the same 13 personal pronouns. It is not until 1960 that we note a drastic drop in this number from 13 to just 1, apparently due to a post-1945 puristic movement.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־-102 (24) Matthew 3:11

ego quidem baptizmo vos in aqua . . ja drje kfću was z wodu (25) Matthew 2:13

qui cum recessissent .. Hdyž bëchu woni woteSli (26) Matthew 2:5

at illi dixerunt ei woni pak prajachu jemu

We have already seen Jordan1s account of overt subject pronoun use in Sorbian in which personal pronouns are only used in emphatic contexts, as is typical for Czech. Only ten years earlier, however, Seiler states the complete opposite:

Die persönlichen Fürwörter werden zum Zeitworte erfordert ... (Seiler 1830: 113) .20

He then lists rules by which personal pronouns are omitted:

Gebraucht und wiederholt werden sie nicht: a) wenn das Subjekt oder sein Pronomen im vorhergehenden Satze .... genannt worden ist ... b) im Imperativ und in Gemeinplätzen, wo man eine bestimmte Person geradezu nicht meint, (ibid)21

'The personal pronouns are required with the verb.' 21 'They [personal pronouns, ML] are not used and repeated: a) if the subject or its pronoun ... was mentioned ... in the previous sentence ... b) in the imperative and in general meaning where one does not refer to a certain person. '

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access It seems that Seiler's description is certainly closer to the real situation found in spoken Sorbian today. It also gives a more realistic account of the use of pronouns in the Sorbian language of his time. For instance, in the preface to his new political newspaper Serbski powëdar a kurër, first published in 1809, J. B. Dejka uses 10 (поп-emphatic) overt subject pronouns (5 x ja; 2 x wy; 3 x my) and only uses a zero pronoun twice.22

The influence of the Czech language on the repudiation of overt pronouns in Sorbian is even more likely when one takes into consideration the fact that three of the authors of the aforementioned grammars studied and worked in Prague at the Serbski seminar (1Sorbian Seminary1) at the time they wrote their grammars, i.e. Jordan (1831-38), Liebsch (1880-83) and Krai (1879-89)- Lusčanski, the co-author of the 1896 Bible, was head of the Serbski seminar in Prague from 1877 to 1694.

4.6. Why 18 Sorbian not a canonical null-subject language? The conclusion that Upper Sorbian is not a pro-drop language in the classic sense raises the question of why Sorbian is different from canonical null-subject languages, such as Czech and Polish. As we have seen in (1), the verbal paradigm of Sorbian is morphologically uniform, and null subjects should

In Lorenz (1981: 106).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access thus be licensed.23 So from the perspective of a licensing/identification approach to the problem the following question presents itself: why is the identification of referential subjects in Sorbian blocked and Sorbian thus prevented from being a canonical null-subject language? Two possible analyses are presented below. These are then followed by a discussion of the limitations of a "pure" licensing/identification approach to the problem and a presentation of a third, alternative analysis.

4.6.1. Identification and the number of syncretisms A possible explanation for why Sorbian is not a null-subject language might be formulated in terms of Roberts's (1993) restriction on the number of syncretisms tolerable for sue- cessful identification. Roberts adopts Rizzi*s (1986) licensing criterion for null subjects:

(27) If Xo is the licensing head of pro, Xo can assign Case to the position occupied by pro.

As pointed out by Roberts (1993: 18), it has been proposed that structural case can be assigned in two different configura- tions, government or agreement, as illustrated below.

23 The issue of overt, non-referential pronouns will be discussed in Chapter 5.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־-105 (28) Case assignment under government: a. X ‘ b. X' X NP Xo YP

NP'

(29) Case assignment under agreement XP NP X 1 V i Xo

There is assumed to be a nominative assignment parameter, according to which nominative can be assigned under government

or under agreement. Nominative assignment under government presupposes movement of the inflected Verb into С position to produce the configuration shown in (28b). This is typical for V-2 languages, such as German, as illustrated in (30).u

(30) Peter liest den Roman. *Peter reads the novel* CP Spec C'

I c — ~ 1P ♦ * . ę

I NP ^ I ' VP I

; i .

ז • : • ז Peterk liestį tk den Roman tį tj

See fn. 13 in Chapter 2 for a possible explanation

of why German is not a pro-drop language. Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־106־ Nominative assignment under agreement, on the other hand, is normal spec-head agreement between AGR as the head and pro in the specifier position of AgrP.

The identification of referential pro {which Roberts calls "content licensing") can happen in two different ways: by "functional richness" and by "formal richness". The pre- condition on formal richness, according to Roberts, is a morphologically uniform paradigm where an agreement slot is available throughout.25 Functional richness, on the other hand, is defined in terms of a restriction on the number of syncretisms tolerable for successful identification, and it allows for one syncretism in the paradigm. Functional richness, as opposed to formal richness, is also compatible with a zero ending (and since the zero ending can be the syncretism, there is actually the possibility of two zero endings in a functionally rich paradigm as long as all the other endings are distinct). We have discussed above how the Sorbian paradigm, and in fact the verbal paradigms of all Slavic languages, fulfill the morphological uniformity condition, which is treated as the licensing condition within the Jaeggli and Safir system, as opposed to Roberts, for whom it is one of the two possible conditions for identification. If we want Roberts's system to

25 In contrast to Jaeggli and Safir's morphological uniformity hypotheses, Roberts's formal richness pertains only to paradigms with consistently derived inflectional forms and not to languages which do not have agreement morphology.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-107-

work we must find a reason why formal richness, i.e. morphological uniformity, is not relevant for identification in Slavic.

It is crucial that a language may not automatically avail itself of either of the two mechanisms of identification. In this context, Roberts (1993: 207) proposes the following general principle, (mainly to account for the loss of preverbal pro in French after a change in the nominative assignment parameter as summarized below).

(31) If Agr° formally licenses pro only under agreement, then Agr° must be functionally rich.

Roberts's theory is based on the history of French, which

once had null subjects, even though pro was only found in certain syntactic environments, unlike in modern Polish or Ital- ian, for example. He shows how the loss of null subjects in French can be linked to the change of a single parameter, namely the nominative case assignment parameter, in conjunction with principle (31) . Old French was a null-subject lan- диаде in the sense that referential null subjects were possible in V-2 contexts, i.e. postverbally under government by AGR, which also means that nominative was assigned under government. Identification in early Old French was possible because the verbal paradigm was functionally rich. A major restructuring took place in the 12th century, and as a result the paradigm ceased to be functionally rich due to the large

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-108- number of syncretisms. However, at the same time the verbal paradigm became formally rich instead. Thus, despite the loss of functional richness, French remained a null-subject Ian- диаде for a time, as long as nominative assignment under government was available. Middle French saw a change with respect to the occurrence of null subjects in that the number of contexts in which null subjects were allowed actually increased -- specifically null subjects could also occur prever- bally. According to Roberts this was possible because nominative assignment was no longer restricted to government config- urations, but instead expanded to include the agreement context as well. Middle French had a morphologically uniform paradigm and content licensing was accomplished via formal richness. The relatively sudden disappearance of null subjects in the 17th century is attributed by Roberts to the fact that nominative assignment became restricted to the agreement configuration, i.e. nominative cannot be assigned under government. At this point, post-verbal null subjects are directly excluded since they cannot be licensed anymore. What about pre-verbal null subjects though? According to Roberts, as a consequence of the change in the nominative assignment pararne- ter, null subjects in a language like 17th-century French can only be identified by a functionally rich paradigm, following principle (31). Formal richness does not suffice any more.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access -109- However, the verbal paradigm of French does not fulfill the functional richness criterion and identification fails.26

In order for Roberts,s approach to be at all applicable to identification in Sorbian we have to make the assumption that the nominative assignment parameter is set to "agreement only".27 In other words pro, is formally licensed under agreement only, and for referential subjects to be identified we need a functionally rich paradigm, while formal richness (i.e. morphological uniformity) is not applicable. A toler- ance of one syncretism for functional richness, as defined by Roberts, may indeed be necessary to incorporate some Slavic null-subject languages. For example, there is a syncretism in Bulgarian between 2nd and 3rd person singular endings in both the imperfect and the aorist, as illustrated in (32) and (33) .

Ä Sprouse and Vance (1994) point out a rather counterintuitive point of Roberts's analysis. Middle French was a language which had nominative assignment both by agreement and by government. However, assuming the validity of princi- pie (31) and taking into account that the paradigm was only formally rich, null subjects in the agreement configuration could only be licensed because of the instances where pro was licensed in the government configuration. 27 This assumption is far from universally held. At first glance one might try an analysis, according to which Slavic languages have different settings for the nominative assignment parameter and thus different ways of identification available to them. However, the setting of the nominative assignment parameter makes no difference for the canonical null-subject languages, since they all seem to have both formally and functionally rich paradigms. For Russian and Sorbian, on the other hand, the setting of this parameter is crucial. King (1993) proposes nominative assignment under government for Russian. However, this would make formal richness relevant for identification, with the undesirable result that Russian should be a canonical null-subject lan- диаде.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access ־110־ (32) jadeše Imperfect: 'eat* 2nd/3rd sg (33) jade Aorist: ,eat' 2nd/3rd sg

Note that this situation does not count as two syncretisms, presumably, because they occur in different tenses. To accommodate these data we would need to revise Roberts's syncretism criterion slightly:

(34) A verbal paradigm is functionally rich if there is no more than one syncretism per tense.

Somewhat more problematic might be the situation in Old Church Slavonic, where one finds two syncretisms within a single tense. In OCS both the imperfect and the aorist dis- play, in addition to the syncretism just mentioned for Bulgar- ian, a syncretism between the 2nd person plural and 3rd person dual, as illustrated in (36) and (38).

(35) jadëaâe Imperfect: *eat' 2nd/3rd sg (36) jadëaâete Imperfect: 'eat' 2nd pl or 3rd dual (37) jastb Aorist: ,eat1 2nd/3rd sg (38) jaste Aorist: ,eat' 2nd pi or 3rd dual

It can be shown (cf. Lindseth and Franks 1994; Franks 1994) that the use of the same form for the 2nd person plural and 3rd person dual does not count as a syncretism in the relevant sense. A syncretism reflects a single form that is non-distinct from more than one feature complex, i.e. it is a defective feature matrix. Verb forms that are lexically associated

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-111- with such defective feature matrices can be substituted for multiple, fully specified AGR matrices at S-structure. A form, such as (35) jadšaše 1 (you-sg/ (s) he) was eating1, for instance, can be analyzed as having the pronominal features lst,+sg] , rendering it non-distinct from both 2nd and 3rd־] person singular agreement. However, when the features cut across both person and number, as in (36) jadēašete 1 (you- dual) was eating* , this is not an instance of feature־p l / (s) he

neutralization, because it cannot be expressed using a single entry in the lexicon. The fact that this is technically not

a syncretism provides a possible explanation for why identification of null subjects is not blocked in OCS. Now compare the Upper Sorbian forms below with the OCS forms in (35-38):

(39) piješe Imperfect: *drink' 2nd/3rd sg (40) piještaj Imperfect: *drink' 2nd/3rd dual (41) wupi Aorist: 'drink' 2nd/3rd sg (42) wupiâtaj Aorist: *drink' 2nd/3rd dual

In contrast to OCS, Upper Sorbian exhibits two real syncretisms within a single tense, violating condition (31) for the identification by functional richness. Common Slavic had distinct forms for the 2nd and 3rd person dual. In (Upper) Sorbian there is no distinction between the 2nd and 3rd person dual (in addition to the inherited syncretism between the 2nd and 3rd person singular in the imperfect and aorist) . The two

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-112- different forms of the Common Slavic 2nd and 3rd person dual have been reinterpreted as a gender distinction, at least as far as the literary language is concerned.

Table 3: The Dual in OCS and in Upper Sorbian OCS Upper Sorbian

Dual Present let nesevë njesemoj 2nd neseta njesetaj/njesetej 3rd nesete njesetaj/njesetej

In literary Upper Sorbian the -taj ending is used in reference to male persons, while -tej is used for all other genders. However, most Upper Sorbian dialects do not express this distinction and use only the ending -tej. We have seen that an approach to identification in terms

of Roberts's restrictions on the number of allowable syncretisms seems to provide an account for the lack of identification in Upper Sorbian.u However, there is one major drawback to this sort of morphological explanation: This account makes the bizarre prediction that the loss of the dual in Upper Sorbian should reinstate the successful identification of null subjects in Upper Sorbian. The Sorbian verbal paradigm without the dual would be "functionally rich" in the

M Roberts's account is intended to explain why French lost null subjects entirely. A strict interpretation of Roberts would thus predict that Sorbian has no null subjects at all, and could not account for the apparent optionality of subject pronouns. In this context see also 4.6.3. for a discussion of the general challenges the Sorbian situation provides for identification/licensing frameworks, and 4.6.4. for an alternative analysis.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-113- sense defined by Roberts and thus no violation of principle (31) would occur.

Ideally, we should compare data from Old Sorbian before

and after the merger of the 2nd and 3rd person dual to see whether this change brought about a concomitant increase in the use of overt subject pronouns. Unfortunately, this is not possible, because the earliest Sorbian texts are relatively recent (16th century) and already show an extensive use of overt personal pronouns in subject position, which is similar

to today's colloquial language. There might be yet another way to determine the validity of Roberts's syncretism account for Sorbian- Some dialects of Upper Sorbian have only infre- quent use of the dual number. Fieldwork might be able to

determine whether or not there are dialects which lost the dual completely, and whether or not such dialects really have the relevant null-subject characteristics which were found to be missing in the dialects which preserved the dual.

In this context, let us return to Russian. On the basis of the null-subject criteria I argued that Russian is not a canonical pro-drop language. An explanation for the lack of identification of referential null pronouns was given in terms of the lack of AGRp^, in the past tense. We might now try to

see whether Roberts,s theory can be applied to Russian as well. It turns out that "defects" in the past tense paradigm can once again be interpreted as the culprit for blocking identification. Russian past tense verbs are marked for gen-

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-114- der in the singular, but this opposition is completely neu- tralized in the plural.

Table 4: Russian Past Tense

Sg. Maec. byl PI. byli Fem. byla Neuter było

Since the singular distinguishes masculine, feminine and neuter, there must be a double syncretism of gender features

in the plural. Identification by functional richness thereby fails.

4.6.2. A word order account While it is in principle possible to formulate an explanation

for why Upper Sorbian is not a null-subject language in terms of a restriction on the number of syncretisms in the morphological marking, I also pointed out the counterintuitive predictions associated with this approach. The non-pro-drop properties of Sorbian may also be at- tributable to some other aspect of its syntax. An inspection of Sorbian texts reveals a striking preponderance of verb- final declarative sentences, in contrast to other Slavic languages, such as Czech, for example. The sentences below are taken from Michałk (1962b) and are drawn from both literary and dialectal sources.29 Sentences (43) and (44) are examples

29 Some of these sentences contain non-standard forms.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access -115- of verb-final subordinate clauses; (45-49) illustrate the verb-final position in main clauses.

(43) nan sebi mysleše, 2 0 so na druhi dźeń father self thought that refi at other day z naâim susodom wróći with our neighbor will-return 'Father thought to himself that he would return with our neighbor on the next day* (44) Marka so woprawdźe wjeseli, zo jónu skoro zaso Marka refi really rejoices that once soon again do Lipska poj&dfe to Leipzig will-drive 1Marka is really glad, that she will soon once again go to Leipzig* (45) a hi20 psaj kaž zło njedobro pfez and already dogs^ like evil not-good through rëpu na Janãowu ćirjeStaj beets at Jančowa r a n ^ 1and already the two dogs came running like the devil through the beets towards JanCowa* (46) tutón jeho pak na puć к śibjeńcy dowjedfe that^ him.* however at way to gallows will-lead *but that one will lead him on the way to the gallcws*

(47) ta husto po Marku na spëwanje pfiûdfe thatfan often for Marka at singing comes *she often picks up Marka to go singing*

(48) mój z Marku najwajacy na polu stojimoj weA»*1 with Mark most-often at field standų 'I and Marka stand the most on the field*

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access (49) najradšo so ludźo njedźelu vrobŽlokaja30 most-gladly refi people Sunday get-drunk 'people prefer to get drunk on Sunday1

In other Slavic languages the sentence-final position may, of course, also be occupied by the finite verb, especially if the verb is the rheme, i.e. the element which provides the newest, most important part of information. In general, the final position in those languages seems to be reserved for the rheme. However, in the Sorbian examples above consider-

ations of functional sentence perspective should not have contributed to the final position of the verbs. What is important for our discussion is that the unmarked position for the finite verb in both the literary and colloquial language seems to be clause-final. Michałk (1962a; 1962b) provides some interesting figures. The ratio of verb-final versus verb-internal position came out to be 62% : 38%. According to Michałk, in 55% of all analyzed sentences with the finite verb in final position the verb did not belong to the rheme. In 86% of all cases where the finite verb did not show up in final position, the verb was immediately followed by the rheme in final position. This indicates that the verb can be dis- placed from its unmarked final position in order to put the rheme there instead. The tendency to put the finite verb into final position, even if it is not the core of the new informa-

Most likely intended to be wożlokaja.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access -117- tion, was found to be even stronger in the literary language than in colloquial speech.

As far as subordinate clauses are concerned, this structure is, of course, very reminiscent of German, where the finite verb must always be in final position.31 In Sorbian this verb-final tendency is true for both subordinate clauses and main clauses, though it is just a preference and not an ironclad rule. In German, the finite verb must be in the second position in main clauses. The Sorbian verb can occur in non-final position, especially when some element other than the verb is clearly the rheme. Based on this discussion I propose that the underlying word order in Sorbian is SOV. For German, SOV order is often derived by analyzing AGRP (and VP) as right-headed, so that when V raises to AGR it will end up in clause-final position, as shown in the structure (50) below?

Even in subordinate clauses, Sorbian word order is not identical to that of German. In the perfect and past perfect, the auxiliary typically follows in second position right after the conjunction (in 87% of all cases according Michałk 1970: 23), while the 1-participle is in final position. Compare the Sorbian and German examples below. (i) hdyž sym! ja z chórom w B. byla2 (ii) wenn ich mit dem Chor in B. geweeen2 bin! ,when I have been in В. with the choir'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־-118 (50) AGRP

יNP AGR VP AGR

NP V

S O ti Vè

I propose that the right-headedness of AGRP also accounts for

word order in Sorbian, the difference being that since AGR

(unless it is a clitic) does not raise to С in Upper Sorbian, conjugated verbs are final in both main and embedded clauses. In analytical tenses, such as the perfect and past per- feet, the auxiliary is typically in second position, while the 1-participle occupies the final position. Consider the following examples, taken again from Michałk (1962b):

(51) ja вут nëkotru zakrasnjenu myš popanydła32 I auX|t| some nice mouse caught 'I caught some nice mice1 (52) mój nan bë tydźeń prjedy wotroőka my father aux3sf1nveff week ago farmhand n&swarił scolded ,my father had scolded the farmhand a week ago1 (53) mi so zda, zo sym prawje wulku rybu pop&nyl me refi seems that auxU| really big fish caught

'it seems to me that I caught a really big fish*

This is very reminiscent of German matric clauses: Ich habe einige Mäuse gefangen. ,I've caught some mice*

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

119-־ Referring back to the proposal in Chapter 3, I assume the existence of different AGR projections (AGR^ and AGR^) in

Slavic. The examples above seem to indicate that all fune- tional projections of the verb except for CP are right-headed. ע:(This is illustrated in the structure in (62

(54) CP Spec С 1

с a g r ^ p

AGIW '׳ N P

A G R ^ P A G R ^

AGIW

TP AGR fend T' / VP I V' NP V ja! syro* t j myš tj tj popanydlaj tk

It is thus possible that the lack of identification of null subjects in Sorbian is somehow connected to the direction

33 If the auxiliary is negated, it loses its clitic status and is found most often in final position (cf. Jenč 1959: 27-28), as illustrated in the example below. Presumably it then occupies the head position of AGR^ or NegP. ja рак jeho widźał njejeym I however him saw not-auxU| *I, however, didn't see him'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access AGRS) . I speculate that the problem in) ,״of branching of AGRp

Upper Sorbian may have to do with the non-adjacency of the head of AGRP (AGR^P), (which would be the AGR carrying the

pronominal features necessary for identification), and the specifier, (which would be the position for pro). This non- adjacency, which makes AGR, despite its richness, somehow not accessible to the null subject, is underlying in all sentences, and is also usually present at S-structure, unless a clitic is present as shown in structure (54).M It is possible that Sorbian developed its predilection for the final

position of the finite verb under the influence of German

subordinate clauses. However, Sorbian has generalized that

structure in its own way, and the problem for the identification of referential subject pronouns can thus be found within the Sorbian syntax itself. Just as with the syncretism approach to the issue of null subjects (or lack thereof) in Sorbian, there are also some potential problems for the word-order based account. The literary language is just as verb-final as the colloquial lan- диаде, if not even more consistently so. However, we have also seen that in literary Sorbian, overt subjects are usually omitted. This situation raises the question of how these

* Note that German, despite its uniform realization -is also not a canonical null ,,״of a relatively rich AGRp subject language. One might speculate that in German too, the fact that thematic null subjects cannot be identified has something to do with the non-adjacency of the head of AGRP and the specifier position (but cf. also fn. 14, Chapter 2) .

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access frequently occurring null subjects are identified in literary Sorbian if identification is blocked by non-adjacency of AGR^ and the subject in [Spec, AGR^P] . It is also the case that there are many instances of null subjects in colloquial Sorbian. If one denies the possibility of identification these should be treated as discourse-licensed ellipses. That would imply that null subjects in literary Sorbian and in colloquial Sorbian are instances of different phenomena, which does not seem very likely. Finally, as pointed out in Lindseth and Franks (1994), one obvious problem with implicating the direction of branching in blocking identification is that there are SOV null-subject languages, which raises the question of why adjacency should be a factor in Sorbian or German, but not in Turkish or Hindu.

4.6.3• Challengee for the identification/licensing framework The problems mentioned in the previous sections arise when we attempt to solve the issue of pro-drop in terms of a highly restricted framework of syntax, trying to account for linguistic phenomena in terms of setting binary parameters and for- malizing precise conditions for the occurrence of certain elements. The paradox is that any non-linguist is able to see why it is possible to omit subject pronouns in both literary and colloquial Sorbian: They are redundant because the verbal endings carry overtly the information which is necessary to recover the pronominal content of the missing pronouns.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

122-־ In general, it seems that most versions of pro licensing/identification, such as Rizzi (1986), Jaeggli and Safir (1989), Rohrbacher (1995), Speas (1994), and also the ones discussed in the previous two sections (Roberts (1993) and the word-order account) will encounter two major challenges with respect to the Sorbian data.35 First, such an approach cannot account for the optionality of overt and null-pronominal subjects in Sorbian. Most versions of the licensing/identification framework predict that a language will either license/

identify null subjects or it will not. Second and related to the first problem, they cannot predict the difference in frequency of null or overt pronouns with different persons of the

verb, especially if such a frequency difference cannot be attributed to functional compensation for missing morphological marking on the verb. We have seen that literary Sorbian does comply with the "subject-pronoun-prohibition rule" given in puristic grammars.

An analysis à la Speas fails to account for the whole range of Sorbian data, although at first glance it might appear to be wide enough to cover systems like Sorbian (though certainly not in a very interesting way). Speas considers "strong" agreement as a necessary, but not sufficient condition to treat agreement affixes as individual lexical items. Sorbian fulfills the conditions on "strong" agreement, and should thus have the option of having lexically listed affixes. Under Speas's analysis, Sorbian has cho- sen otherwise, by listing agreement affixes as part of the verbal paradigm. However, this should make the occurence of any null subjects a violation of the principle of economy, and we cannot account for the optional character of null subjects. Moreover, this type of analysis would also fail to give an explanation for why Sorbian is different, and instead blame it all on a "lexical idiosyncracy" of the Sorbian agreement morphemes.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־123־ especially with respect to 1st and 2nd person subject pronouns. However, noneraphatic 3rd person pronouns do occur more frequently in literary Sorbian.36 The following are examples of поп-emphatic pronoun subjects from the novel by Koch (1991) which was mentioned above for its zero subjects:

(55) a. Rēč zetkanja zmóSni. Wona teŽ language meetings will-enable it too powitańje pfewozmje greeting will-take-over ,The language will make meetings possible. It will also serve to greet.'

b. Spisowaćeljo, kotfiž so w pfidatnej rēči writers who refi in another language wuznaja su w pfidatnym swSće žiwi. Woni su know aux^in another world live they aux^ pola susoda doma, at neighbor at home 1Writers who know an additional language, live in an additional world. They are at home at their neighbor's.1

Also recall the literary example (21a) repeated here as (56), which shows that even in the literary language a personal pronoun may be coreferential with the matrix subject.

(56) Feliks njepytny, zo wón hižo hodźinu po mēsče honja ' F. didn't realize, that he had already been running through town for an hour*

Ä Steven Franks (personal communication) reports a similar phenomenon for Bulgarian.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access A particularly striking (and apropos) example is the following text found in the Šewc grammar (1974: 118) . Bear in mind that this is the same grammar which states that there is only a very limited use of personal pronouns in Upper Sorbian.37

Personalne pronomeny wón, wona . . . njepoćahuja so . . . jenož na wosoby, ale tež na zwërjata a wëcy. Wone maja pïez to zdobom šērši woznam. Wosebitosć personalnych pronomenow 3. wos. zwuraznja so dale w tym, zo poćahuja so wone zwjetśa na nēšto, štož je hižo z konteksta znate. Wone wobhladuja so tohodla jako wosebita skupina pronomenow ... 1The personal pronouns wón, wona ...('h e 1, ,she' . ..) do not only refer to people, but also to animals and things. Therefore they have a broader meaning. The special character of the 3rd person pronouns finds further expression in the fact that they more often than not refer to something which is already known from the context. They are therefore considered a special group of pronouns...'

This skewing in person appears also to be reflected in colloquial speech. Even speakers who are influenced by the literary language and tend to omit the personal pronouns of the 1st and 2nd person will frequently use a 3rd person pronoun. Consider the following examples:

(57) (ja) bydłu w BudyŠinje 'I live in Bautzen' (58) woni bydła w Budyšinje 'they live in Bautzen'

37 To be fair I need to mention that Šewc (1984: 117) only refers to 1st and 2nd person pronouns in saying that their use is reserved to emphasize the subject. On the other hand, his comment about the restricted use of pronouns refers to all persons, and he does not imply anywhere that the pronouns of the 3rd person are used more frequently. Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־-125 (59) wón bydli w Budyšinje 'he lives in Bautzen1

In (57), about 70% of my informants38 used overt ja, while 100% used woni in (58) and wón in (59) . This was true in the per- feet tense as well:

(60) (ja) sym sej toäu kupił I аихІЦ refi bag bought^ II have bought myself a bag1 (61) wón je sej tošu kupił he aux^ refi bag bought^ ,he has bought himself a bag1

Even speakers who omit the personal pronoun in (60) use wón in (61). In fact, most native speakers felt that the personal

pronoun of the 3rd person in sentence-initial position was

obligatory. When shown literary examples with 3rd person null subjects in initial position they felt that these were "unusual" and "misleading".39

38 I interviewed approximately 20 native speakers of various ages and educational backgrounds. 39 Virtually all informants also used overt 3rd person pronouns in other positions in the sentence, as shown in (i) and (ii) . However, most of them also seemed somewhat more willing to accept constructed examples without the pronoun. (i) je sej wón tu tošu kupił aux3sf refi he this bag bought 'did he buy the bag yesterday' (i1) ? je sej tu tošu kupił (ii) wčera je sej wón tu tošu kupił ,yesterday he bought the bag' (ii1)? wãera je sej tu tošu kupił

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access The question arises whether there might be a connection between this difference in the frequency of overt pronouns and the verbal morphology of Sorbian. It is true that in the 3rd person present tense, as well as in the imperfect and past, there is no marker on the verb which specifies the gender so that the use of a personal pronoun might be blamed on the otherwise threatening ambiguity. In other words, kupuje (1buys') can refer both to wón and wona, just as the plural kupuja can refer to wone or woni. However, if this were the reason, then we could expect frequent omission of the subject pronoun in the 3rd person singular in the perfect tense, where the gender is unambiguously expressed on the I-participle. This expectation is not borne out, as the comparison of exam- pies (59) and (60) shows. Moreover, in other Slavic languages which are pro-drop languages, such as Czech and Polish, the same situation obtains as far as the non-expression of gender in the present tense of the 3rd person is concerned. Never- theless, in those languages null 3rd person pronominais are totally normal in unmarked contexts. In this connection I argued in 3.2. that only AGR^, and not AGR^, seems to be relevant for the identification of null subjects.

4.6.4. The competition model and the loss o£ pro-drop Most versions of the licensing/identification framework follow the traditional wisdom that there is a direct link between the inflectional richness of a language and the availability of

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access null pronouns. While it seems plausible that a language would lose the ability to have null subjects if it loses the morphological marking of agreement features on the verb, there are languages, such as Icelandic, German, French, Rhaeto-Romance of Switzerland and some northern Italian dialects (cf. Oreänik (1986: 266)) which resist such a purely functional view of the loss of null pronouns.40 Orešnik points out that these languages developed obligatory subject pronouns when they pos- sessed (and some still possess) full and unambiguous morpho• logical markings on the verb.

Sprouse and Vance (1995) provide an interesting alternative to attempts made strictly within the principles and parameters framework to account for the (gradual) loss of null pronouns in languages which were once pro-drop languages.

Based on data from the history of French and the Surselvan dialect of Rhaeto-Romance they argue against a purely fune- tional view of the loss of pro-drop in terms of a change (decline) of morphological marking on the verb. Instead, following work by Kroch (1989), they show how a competition model of

Orešnik (1986) points out that in Old High German there was a person skewing with respect to the use of null- subject pronouns. The pronominal subjects were more frequent in the first and second persons than in the third. He asso- ciates the origin of unemphatic pronoun subjects with a ”strengthening", which started first in the so-called "strong" positions. The third person, as the least marked person is also the least likely to become strengthened. Orešnik thus proposes that the same state of affairs with respect to the differences among the verbal persons should obtain in all languages. It seems that Sorbian is a counterexample to this hypothesis, because we find a clearly higher frequency of 3rd person pronouns in that language.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00062018

-128- language change can account for a syntactic change over time, one example of this being the gradual loss of pro-drop properties in a language. I will not present the details of Sprouse and Vance's analysis here, but the following is a short summary of their hypothesis. Two forms of a language are said to be in "competition" if they make exactly the same contribution to the semantic meaning. Once a language has two competing forms, speakers attempt to match the relative frequency of these forms. Over time the relative rates for two competing forms will shift in favor of the form which is parsed more reliably, whereby "parsing" simply means grammatically based understanding of an utterance. Once this process is set in motion it will finally lead to the predominance of one form and the loss of the other. Applying this idea to the issue of null subjects, it can be assumed that sentences with overt pronouns would be parsed more successfully than sentences with null pronouns. However, in order for a drift towards a high frequency of overt pronouns to take place, null and overt pronouns in that particular language must have the same referential properties. Oth- erwise, they will not be in competition and the pro-drop character of the language will remain stable. Consider the following Czech sentences:

(62) a. ja mluvim česky 'I speak Czech1 b. e mluvim česky. 'I speak Czech'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access The overt pronoun and the zero form in the two sentences above do not represent competing forms. Ja in Czech is a tonic pronoun and is used emphatically. The zero pronoun, on the other hand, is an atonic pronoun. The different referential properties of overt and zero pronouns in a null-subject lan- диаде such as Czech can also be illustrated in the following type of construction:41

(63) a. PetTj myslí, 2e *on*/pro, má mnoho penëz ,Peter thinks that he has a lot of money1

As we have already seen in the course of the application of the null-subject criteria to Czech, the overt pronoun cannot be bound by the higher subject in the matrix clause, even though this is possible for the zero pronoun. In other words, while pro seems to be subject to condition B, the overt pronoun in a null-subject language seems to obey binding condition С (which boils down to the same relevant difference in the referential properties between overt and null pronouns previously pointed out by Jaeggli (1986) and Montalbelli (1984) ) .

In Sorbian, on the other hand, the situation is actually quite different. We demonstrated above that overt pronouns are not restricted to emphatic contexts and that embedded

Sprouse and Vance propose the same type of construction as a reliable criterion to identify competing forms.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access overt pronouns can be coreferential with a matrix subject. Example (21f), repeated here as (64), illustrates this.

(64) Jana! je prajiła, zo chce wona./в, rady kupać hic 1Jana, said that sheį would like to go swimming1

Thus there is a true competition in the relevant sense between overt and zero pronouns. If we define a null-subject language in terms of the three diagnostics in (20), a language ceases to be pro-drop as soon as there are atonic overt pronouns. The advantage of the competition model is that it allows for a gradual change in the frequency of null subjects over time, including a period of optional pro-drop, without being pressed to look for a change in a parameter setting. The competition model is not concerned with the formulation of precise licensing and identification conditions, though it does not explicitly deny their existence. For this sort of approach it is totally expected that there are languages in transition, during which there are both zero and overt subjects. For a licensing/identification approach there is no in-between: either null subjects are licensed or they are not. Sprouse and Vance point out that their theory is not incompatible with a parameter-setting approach such as the one presented by Roberts (1993). Their competition model is best equipped to account for a gradual loss of pro-drop properties without concomitant change in the explicitness of verbal morphology. By itself though, this model fails to explain any

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access sort of sudden change, such as the disappearance of post-verbal pro in French during the 17th century discussed in section 4.5.1. Applied to Sorbian, the competition model of language change might give the following account for the loss of canonical null-subject properties. At an early stage atonic pronouns entered the Upper Sorbian language. The source for this is not clear, but German influence could certainly have played a major role as the region became bilingual.42 Another possible origin of atonic personal pronouns might be found in the relative order of the auxiliary and the participle in the perfect and past perfect. In contrast to Czech or Polish, in Sorbian the auxiliary has to precede the 1-participle in neutral style. In Czech, the order vidël jsem (1seen havellf ') is normal and in fact frequently found in the initial position of a sentence. The Czech auxiliaries are second-position clitics, and the form jsem in vidël jsem is clitized to the preceding participle. Similarly, Sorbian auxiliaries are enclitics.43 However, for Sorbian the word is not available.44 The ״participle - auxiliary״ order option auxiliary has to precede the participle, so that ja sym widźał,

Sprouse and Vance do not identify the origin of atonic pronouns in Romance languages, but they doubt that a neighboring language had any role in that process. 43 According to Michałk (1965), Sorbian clitics prefer the "Wackernagel" position, although post-verbal position can exist if the verb occupies the second position in the clause. 44 This order is possible if the auxiliary is negated (cf. fn. 33).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access with the overt personal pronoun preceding the clitic, is the equivalent to the previously mentioned Czech example vidël jsem. A Czech speaker would not use the same order of «le- ments as Sorbian, unless the subject is supposed to be especially emphasized. As discussed in section 4.4, it can be assumed that by the 18th century overt pronouns had established their predominance over the null variant, both in spoken and written Sor- bian. However, in the second half of the 19th century the frequency of null pronouns suddenly increased again, mainl7 in the written language, but presumably also in spoken Sorb;an.

Null pronouns were artificially (and forcefully) favored by prescriptive grammars trying to root out the German influence in the Sorbian language and recreate Sorbian as a real m.11- subject language based on the Czech model. Such a shift in the literary language may have made the pro-drop phenomenon socially significant by introducing what we might call a prestige factor" which artificially influenced the relative״ frequency of overt and null pronouns.45 The better educated and well-read members of the population began to try to imitate the literary language by reducing the frequency of overt pronouns in their speech. This tradition is still alive to*

5 Sprouse and Vance mention that they abstract cway from prestige factors in the history of French null subjects. In the 17th century French grammarians discouraged the orris- sion of pronouns, which might have caused an artificial shift in the relative frequency of overt and null pronouns in favor of the latter.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access day, as shown earlier in this chapter. Uneducated, simple peasant speech in literary work is imitated in part by the obvious use of overt pronouns. Moreover, all Sorbian children are taught in school to avoid the (over)use of overt pronouns. This all may explain why Sorbian today still has a competition between overt and null pronouns. Moreover, it can also account for the direction of change (overt pronouns -* null pronouns), which apparently goes in the opposite direction from what the competition model predicts and is particularly striking in literary Sorbian, where null subjects became dominant. Overt pronouns do predominate in colloquial speech; however, the influence and frequency of the zero variant is great enough to keep the competition going. Under such an account Sorbian should be analyzed as a truly optional null-subject language. In most environments null and overt pronouns seem to occur basically in free variation, with preference given to the overt variant. We might even try to reconcile a competition model approach to null subjects with a licensing/identification approach. Sprouse and Vance's model is meant to account for the change of pro- drop properties over time. They do not deal with the problem of what allows or prevents null subjects originally, i.e. they are not concerned with licensing and identification conditions. If we assume that there is some syntactic mechanism in the grammar of a native speaker which ensures the recoverability of null subjects, the question arises of whether or not

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access this mechanism remains intact in the process of a gradual loss of pro-drop. Put differently, we can ask the question of whether or not Sorbian is really a non-pro-drop language, in the sense that identification of referential null subjects is blocked. Perhaps the licensing/identification mechanisms have not been lost in Sorbian, so that there has not been a parametric shift as such. It is thus possible that the occurrence of null subjects in Sorbian (both literary and colloquial) is indeed a case of genuine pro-drop and thus does not have to be analyzed as discourse licensed ellipsis. This approach would eliminate the problems raised at the end of section 4.6.2., specifically the question of how to account for the identification of null subjects in literary Sorbian. It also explains why purism could be so successful in (re)introducing null subjects into the literary language: the syntactic mechanism of identifying null subjects was (and is) intact in the grammar of native speakers. The null-subject diagnostics, which are based on the clear division of labor between null and overt pronouns found in canonical null- subject languages, such as Italian and Spanish, fail due to the atonic character of Sorbian personal pronouns. Interest- ingly, this lack of a division of labor is true not only for the colloquial language, but also for the null-subject literary language, as shown in example (56) above.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-135- 4.7« Lower Sorbian

The main focus of this chapter has been Upper Sorbian. Al- though Lower Sorbian lays claim to a separate literary lan-

диаде, it is clear that the two languages are very closely related. It is not my intention to support one side or the

other in the ongoing debate about whether there are really two

separate Sorbian languages, but it is still worth looking at Lower Sorbian in the context of null-subject phenomena.

Lower Sorbian developed the same ambiguity between the 2nd and 3rd person dual as Upper Sorbian. In the aorist and imperfect paradigms this leads to a double syncretism, since it also preserved the syncretism between the 2nd and 3rd person singular from Common Slavic. Compare the Lower Sorbian verbal paradigm below with that of Upper Sorbian in Table 1.

Table 5: Verbal Paradigm of Lower Sorbian žēlaā *to work* Present/Future Imperfect Aorist

Sg. 1. žēlam žēłach nažēłach 2. žēl aš žSlaSo nažēla 3. žē ta žSlaāo nažSla Dual 1. žēłamej žēlachmej nažēłachmy 2. žēiatej žšlaātej nafSlaštej 3. žēlatej žSlaŠtej nažēlaātej PI 1. žēłamy źółachmy nažēłachmy 2. žēłašo žētaščo nažēiaščo 3. žēłaju žēłachu nažēłachu

According to standard Lower Sorbian grammars, personal pronouns in subject position are not obligatory and usually

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-136- omitted. Starosta (1991:40) gives a description reminiscent of a null subject language. According to him, personal pronouns are only used in emphatic contexts or in order to avoid misunderstandings. Janaá (1984: 171) is more noncommittal, describing their use as optional. In general, the use of overt pronouns in Lower Sorbian seems to be comparable to Upper Sorbian. The following examples are taken from Nowak's (1988) Gutes Niedersorbisch, which is a guide to "good" usage of the Lower Sorbian language.

(65) my stawamy akie ze słyńcom we get-up only with sun ,we only get up with the sun'

(66) ze switanim jo won se wuporał with sunrise aux3s| he refi left 'with sunrise he set off' (67) ga sy ty cora wusnuł when auXfcg you yesterday fell-asleep 'when did you fall asleep yesterday'

(68) ja som poł nocy wocuśeta lafcała I auxlłf half night awake lay 'I lay awake half the night' (69) my wjele nowego njezgonijomy we much new not-hear • we don't hear much new'

Note that in all these examples there would not have been any risk of ambiguity had the pronoun not been used. Neither are these sentences examples for overt pronoun use in emphatic or

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access contractive contexts.46 Thus, Lower Sorbian is not a canonical null-subject language in the sense defined in previous chapters. The application of the other two null-subject criteria leads to the same conclusion. Overt Lower Sorbian woni (,they1) can have arbitrary reading, thus contradicting diagnostic (20c). Similarly, an overt pronoun may have a bound variable reading, causing diagnostic (20b) to fail. This is illustrated in the examples below.

46 Word order in Lower Sorbian has not been studied in as much detail as in Upper Sorbian. In principle the word order pattern seems to conform more or less with that of Upper Sorbian. Waurick (1968: 126) writes:

Wie im Obersorbischen herrscht auch im Nieder-sorbischen der Grundsatz, daß das nichtzusammen- gesetzte verbale Prädikat (verbum finitum) in der Regel am Satzende steht. Jedoch scheint diese Regel etwas großzügiger angewendet zu werden als im Obersorbischen, denn verhältnismäßig häufig tritt Zweit- oder Mittelstellung des nichtzusammen- gesetzten verbalen Prädikats auf. Ähnlich wie im Obersorbischen stehen bei einem zusammengesetzten verbalen Prädikat die finite Form des Hilfverbs gewöhnlich in Zweit- oder Mittelstellung und das ł- Partizip bzw. der Infinitiv am Ende des Satzes. Jedoch gibt es auch hier wahrscheinlich zahl- reichere Ausnahmen als im Obersorbischen. 1As in Upper Sorbian, in Lower Sorbian there is a principle according to which the non-compound verbal predicate (finite verb) usually is in final position. However, this rule seems to be applied less strictly than in Upper Sorbian, because relatively often the non-compound verbal predicate occurs in second or middle position. As in Upper Sorbian, the finite form of the auxiliary in compound verbal predicates is usually in second or middle position, and the /-participle or infinitive at the end of the sentence. However, there are probably again more exceptions than in Upper Sorbian.1

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-138- (70) a. w Londonje woni zasej stajaškuju in London they again strike 1in London they are again on strike* b. we Lužycy woni maja rëdne stare nałogi in Lusatia they have nice old customs *in Lusatia they have nice old customs* (71) a. kuždy nan se mysli, až won ma mudrego syna every father refi, thinks that he has smart son 'every father thinks that he has smart son'

se naźeja, až se won skoro ־b. kuždy pïosaf every beggar refi hope that refi he soon rozbogaśijo becomes-rich *every beggar hopes that he will soon become rich' c. Frido wë, až won njebužo ains krydnuś Frido knows that he not-will A receive 'Frido knows that he won't get an A*

As above with Upper Sorbian, it should be emphasized that there are also many instances of null-subject pronouns in Lower Sorbian, so that the overt and null variants seem to be in free variation. Below are some examples with the null variant, taken from Nowak's (1988) book.

(72) som jogo poznł wot jog nejlēpšego boka auxUj him got-to-know from his best side *I met him from his best side' (73) cas Źywjenja njejsom jogo poznał time life^ not-auxlsg him met 'I have never met him in my entire life*

(74) na dobru gluku smy do kjarcmy zajšli at good luck auxIpl to pub went 'we went to the pub on the off-chance'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־139־ The use of overt pronouns is widespread in the earliest texts, such as the Lower Sorbian Bible translation from 1548

by MiklawuS Jakubica. The translation of the New Testament by Jan Bogumił Fabricius in 1709 shows a very frequent use of overt personal pronouns. A direct comparison of one passage (Luke 10:31-35) reveals six overt pronouns in the Jakubica translation and nine overt pronouns, (including one overt expletive), in the Fabricius translation. The Fabricius translation is cited below.

(75) Luke 10:31-35; Fabricius 1709 31. Wono pak pšifožecy se stanu, až jaden mjerãnikna tej samej droze dołoj žešo, a gaž jogo wupyta, žešo won mimo. 32. A teje rownosći tež lewita, gaž won na to mèsto pšiže, a jogo wižešo, žešo won mimo. 33. Jaden Samariski pak ducy na tej droze, pšiže tam, a gaž won jog wižeŠo, bu jomu jogo luto. 34. A pšistupjecy k njomu hobalowaSo won jogo rany, a lejašo nutš wolej a wino a sajźi jogo na swojo zwërje, wjēžešo jogo do gospody a zažarbowašo jogo. 35. A na drugi źeń dalej ducy wujmje won, a da jej tomu ZaŽarbuj jogo, a״ :gospodarjoju, a žaŠo к njomu bužoš li njeco wjecej nałoźyś, cu ja tebe jo zapłaśiś, gaž ja pšidu".

Similarly, in Matthew 2:2-18, i.e. the same passage we used before for comparison between the different versions of Upper Sorbian Bibles,47 there are nine overt, поп-emphatic pronominal subjects in the Jakubica and ten in the Fabricius version. The use of overt pronouns is also evident in the everyday language of the 19th century, as illustrated in the fol

See section 4.5.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access -140- lowing excerpts taken from Nepila's "Głodne lēto". Nepila had very little formal training in writing and his autobiographical stories are said to closely reflect his native dialect of Lower Sorbian. Because of this, he is considered the first ludowy spisowaćel (1writer of the people1)

... ja cu Warn tež jenu nowu powjesć prjodk ćinjesć wót togo drogego lēta tedym hako ja som mały był a kajku ja jēdž jëdl som tedym we drogem a głodnem lēče ...

' I want to give you a new account of that year, back then when I was little, and what kind of food I ate during that year of increased prices and starvation • • • t

... da wóni su tež wopytowali wót šogo jēsč, ale wóni su jare mało zjēdli; wóni su jēsč wopytowali, ale wóni njejsu mógli teje jēdži jēsč. Wóni wopytowali tež derje su, ale wóni su šycko roztyłja roz- chytali ...

1and they too tried to eat from everything, but they ate very little; they tried to eat, but they couldn't eat that food. They had a taste of it, but they threw everything around ... 1

The serbske narodne wozrodźenje did not reach Lower Lusatia until the end of the 19th century. Even then, the striving for purism in the literary language was not quite as strong as in Upper Sorbian. The version of the Lower Sorbian New Testament published in 1895 in Halle was still Fabricius 's

(1709) translation. In Upper Sorbian, on the other hand, a new Bible translation by Lusčanski and Hórnik was published which was "cleansed" of virtually all subject pronouns which

Examples cited from Lorenz (1981: 94-98).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00062018

־-141 were so abundant in Swëtlik (1690) .49 The Upper Sorbian exam- pie below is the same passage as the Lower Sorbian in example

(75) . There is only one overt personal pronoun in (76), com-

מ.(pared to nine in (75

(76) Luke 10:31-5; Lusčanski and Hórnik 1896 31. Sta pak so, zo mēšnik po tom samym puču džēše, a wuhladawśi jeho, pfeńdźe nimo. 32. Tak tei lewita, jako tam porno pfińdźe a jeho wuhlada, pfeńdźe nimo. 33. Jedyn Samaritan pak tam pućujo pfińdźe porno njemu, a wuhladawSi jeho bu z miłosću hnuty. 34. A pfistupiwśi, zawjaza jeho ramy, wolij a wino na nje linywśi, a sadźi jeho na swoje skoćo, dowjedźe jeho do hospody, a zastra jeho. 35. A nazajtra wucahnywâi dwaj dźesatnikaj da jej hospodarjej a praji: Staraj so za njeho, a Štožkuli wjacy wuložiš, ja, hdyž so wróću, zapłaću tebi.

In summation, it appears that Lower Sorbian behaves more or less like Upper Sorbian with respect to the use of overt and null-subject pronouns.

4.8• Summary

9 In this chapter I discussed the status of Upper Sorbian with respect to null-subject properties. Most grammarians describe Upper Sorbian as a canonical null-subject language, and the literary language in general avoids overt subject pronouns

Also compare section 4.4. 50 The personal pronoun in Luke 10:35 of the Lusčanski and Hórnik version is also present in the Latin Vulgate: Et altera dir protulit duos denarios, et debit stabulario, et ait: Curam illius habe: et quodcumque supererogavis, ego cum rediero reddam tibi.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־142־ (with the possible exception of 3rd person pronouns). However, the situation is quite different in the spoken language. There, nonemphatic overt pronouns occur very frequently. I argued that the gap between literary and colloquial languages is the result of a strong movement for language purism which started in the second half of the 19th century and strove to eliminate elements of German origin from the Sorbian literary language. An application of the three null-subject diagnostics to Upper Sorbian revealed that Upper Sorbian does not show the division of labor typical for a real null-subject

language. I introduced three potential accounts of why Sorbian might be different from other West Slavic languages with respect to its null-subject properties. Two of the accounts were formulated within the licensing

and identification framework. One possible solution to the question of what might be blocking the identification of referential pronouns was explained in terms of Sorbian crossing the threshold of allowable syncretisms required for functional richness of a verbal paradigm. The other possible account for the blockage of identification was derived from the peculiar SOV word order of Sorbian. Sorbian is underlyingly (and in most cases also at S-structure) V-final. I speculated that the problem for identification may have something to do with the non-adjacency of AGR and [Spec,AGRP]. The third proposal was based on an application of the competition model of language change to the issue of the grad

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-143- ual loss of pro-drop, as proposed by Sprouse and Vance (1995) . According to this analysis, a null-subject language will start moving towards the complete loss of pro-drop as soon as atonic overt pronouns enter into competition with null pronouns. Atonic pronouns in Sorbian may have come about under German influence and/or as a result of the relative order of the participle and its auxiliary in the perfect and past perfect tenses.

All three accounts were able to explain why Sorbian is

not a null-subject language in the strict sense (despite the fact that they had difficulties in accounting for certain data and/or made counterintuitive predictions). I argued that any strict application of the licensing/identification framework encounters problems in accounting for the optionality of

Sorbian null/overt subjects. Thus, only the competition model analysis seemed flexible enough to make allowances for the specific character of Sorbian as an optional null-subject language, unless we want to assume that the presence of referential, atonic pronouns and the availability of an identification mechanism are not mutually incompatible.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 5. THE STATUS 07 OVERT EXPLETIVES IN CZECH AND SORBIAN

5.1. Licensing/identification and the typology of languages

In the previous chapters we were mainly concerned with the occurrence or lack of overt referential subject pronouns. However, the lack of overt referential pronouns in a language does not necessarily mean that the language allows no null subjects at all. In the review of the development of the pro- drop parameter in Chapter 2 we saw that the existence of mixed languages (which have null non-referential (expletive) subjects but require overt thematic subjects) has suggested that the null-subject parameter cannot be monolithic as originally assumed.1 As we have seen in Chapter 2, one standard approach to the problem of mixed languages is to capitalize on the distinction between licensing and identification conditions on null subjects. The licensing conditions specify the environment in which null subjects are allowed to occur. All null subjects, independent of their thematic status, must be formally licensed. However, only referential null subjects re-

1 Within government binding theory, expletive subjects are noun phrases in subject position which are not assigned a thematic role (such as agent, patient etc.). They are thus different from referential pronouns on the semantic level. There are also a number of syntactic differences. For example, in contrast to referential pronouns expletives cannot be questioned (except in tag questions) and cannot be coordinated with another NP. This is illustrated in the examples below: (the beer ־ i) It is cold. What is cold? (it) It is raining. *What is raining? (ii) It and the coffee are cold, (it = the meal) *It and the coffee are cold, (with it = expletive)

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-145- quire additional identification to ensure unambiguous recoverability of their pronominal content. Null expletives, on the other hand, merely need to be licensed. In a mixed lan-

диаде pro is formally licensed to occur in a certain position but is not identified. It follows that in such a language pro is restricted to expletives, which, due to their non-referential status, do not need to be identified. This type of system predicts the three different language types in Table 1.

Tabi• 1: Possible language types (I) i. canonical pro-drop languages ii. mixed type: overt referential pronouns; null expletives iii. canonical non-pro-drop languages

The essence of the licensing/identification dichotomy is

that it is much easier for expletives than for referential

pronouns to be null, since their pronominal content does not need to be identified. Identification is more *1difficult" than licensing and can perhaps be understood as **more

complete" 1icensing.2 The licensing/identification framework thus predicts that a language type with null referential and obligatorily overt expletive pronouns cannot exist. Hyams (1986: 71-74) points

See also 2.3. about the practical impossibility of identification without licensing. Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access out that this prediction follows from Chomsky's avoid pronoun ג: (principle (1981: 65

(1) Avoid Pronoun Principle

Avoid lexical pronominal if a null pronominal is possible.

A language which allows null-referential subjects must be able to license null subjects in general. If pro is licensed, then a language will have null expletives. Indeed, Hyams extended this prediction to state that not only will expletives in such a language be allowed to be zero, they must be zero. In other words, languages which license phonologically null subjects will not contain lexical expletives at all. Note that even in a canonical null-subject language the subject pronoun is sometimes lexicalized. However, its use is reserved for special functional purposes, such as emphasis, contrast or change of discourse topic, and also avoidance of ambiguity, and thus is not a violation of the avoid pronoun principle. In certain contexts the overt variant cannot be "avoided", because it carries some additional meaning which the zero variant is lacking. The use of overt subject pronouns in a canonical null-subject language is thus determined by functional considerations. Expletive pronouns, on the other hand, are by definition void of semantic content, and it is a standard assump

3 The status of a language as an optional pro-drop language, as was suggested for Sorbian in section 4.5.4., may jeopardize the general validity of Chomsky's principle. Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

147-־

tion that they cannot be used for functional purposes. Thus, there should never be a reason to use an overt expletive sub-

ject in a null-subject language. Based on this idea, language

acquisition theorists such as Hyams (1987: 74) have argued

that the existence of overt expletives in a language serves as

an unambiguous signal to the language learner that the lan- диаде does not allow empty subjects of any kind. According to Hyams, the presence of overt expletives in a language is a

"trigger" which induces the restructuring of the original drop] which is supposed to be the־parameter setting [+pro unmarked value of the null-subject parameter of UG. Franks (1990) provides an interesting alternative approach to null subjects which places emphasis on different case assignment properties. He proposes two (more or less independent) parameters :

(2) Nominative Case features must be realized. (3) Nominative Case assignment cannot take place at LF.

Descriptively, Franks calls (2) the overt expletive subject parameter [OESP], while (3) is the overt subject parameter

[OSP]. In contrast to the previous system, these parameters do predict four different language types:

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-148- Table 2: Possible Language Types (II) i. [-OSP,-OESP] optionally overt thematic subjects, obligatorily zero expletives ii. [+0SP,-OESP] obligatorily overt thematic subjects, obligatorily zero expletives iii. [+OSP,+OESP] obligatorily overt thematic subjects, obligatorily overt expletives

iv. [-OSP,+0ESP] optionally overt thematic subjects, optionally overt expletives

This notation is somewhat misleading, since [+OSP] means

obligatorily overt thematic subject, but [+OESP] can mean obligatorily overt or optionally overt expletive subjects, depending on the setting for the [OSP] parameter. Similarly,

[-OSP] means optionally overt thematic subjects, while [-OESP] means obligatorily zero expletives. The combination [+OESP] "OSP] is thus not an instantiation of the fourth "forbidden־]

type4 under the licensing/identification approach. Instead, it means that while nominative case must be discharged to subject position there is the option to assign nominative at LF rather than at S-structure. These asymmetries between [OESP] and [OSP] are caused by the interaction of both nominative case parameters in (2) and (3) with the so-called case filter, according to which an NP must have case in order to be

Forbidden type: null referential subjects, obligatorily overt expletives.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-149- visible. Specifically, as mentioned by Franks (1990), there are two sides to this requirement: all overt NPs must be visible at PF; only theta-marked NPs must be visible at LF.

The main focus of this chapter is to discuss how Czech and Sorbian data are compatible or incompatible with the licensing/identification framework. Therefore, I will not follow this interesting line of argument any further here and will come back to Franks's (1990) proposal only as an aside. Below I will show how overt non-referential subject pronouns in

spoken Czech, although in contradiction with Hyams's predic-

tion about the total lack of expletives in any null-subject language, can be predicted by the licensing/ identification system without a violation of Chomsky's avoid pronoun principle. This is followed by a discussion of Upper Sorbian expletive-like

pronouns. Because of the distributional restrictions on the occurrence of such pronouns I will argue that they do not cause any problems for the licensing/identification approach either.

5.2. Spoken Czech5 The standard, written language (spisovná čeština), was codified at the beginning of the 19th century by Josef Dobrovskÿ. Spisovná čeština is based on 16th-century Czech, under the assumption that the German/Habsburg repression after 1618 had corrupted "pure" Czech. This romantic view of historic Czech

5 The hypotheses and data in section 5.2. and 5.3. have, in part, been previously presented in Lindseth and Franks (1994) and Lindseth (1994).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access ignored several changes which had been developing in the Czech language since the 16th century. The written language was in many respects already archaic at the point in time when it was recreated. It is, as the name spisovná indicates, a written and not a spoken language. The spoken language developed in a natural way without the artificial intervention of philolo- gists. Thus there is a historic and linguistic gap between literary (written) Czech and spoken Czech. It is important to understand that the difference between the two cannot be compared to the difference between the written and spoken variants of other languages. As Townsend (1990) points out, all languages have layers of slang, dialect words, jargon. In Czech too, these layers are incorporated in the spoken Ian-

диаде, and various emotive and subjective factors are relevant. However, there are large numbers of neutral pronuncia- tions, morphological forms, syntactic constructions and words used by educated speakers in everyday conversation which differ from those prescribed by the literary language. The term "spoken Czech" is somewhat problematical. It is sometimes used to refer to Common Czech (obecná čeština), which is an interdialect spoken in Bohemia and Southwestern Moravia, or it can refer to Commonly Spoken Czech {bēžnē mluvená čeština), sometimes also referred to as Colloquial Czech. Several linguists (Hammer 1985, Short 1993, Townsend 1990) see the literary language (spisovná čeština) and Common Czech (obecná čeština) as the two basic codes, while Commonly Spoken

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-151- Czech {bēŽnē mluvená čeština) is somewhat of a transitional stratum. Short describes it as a "channel" by which "upwardly mobile features" of Common Czech can enter the standard lan- диаде. Hammer (1985) describes it as a blend of literary and

Common Czech, whereby Common Czech is the foundation, and speakers switch to the literary forms for functional purposes of emphasis, irony, intimacy, etc.

The following discussion is not focused on the relation-

ship between the various forms of the Czech language. In- stead, I will investigate only one syntactic phenomenon which can be found in the spoken language (in both obecná češt ina and bēžnē mluvená češt ina) .

5.2.1. Syntactic statue of overt expletives The Czech personal pronouns are já (1I 1), ty (1you1), on, ona» ono (1he, she, it') in the singular, and my ('we'), vy ('you'), and oni ('they1) in the plural. In stylistically unmarked

discourse, Czech (both literary and spoken) omits unstressed pronominal subjects. This is illustrated in the passage below taken from Heim (1982: 151).

(4) 1. Potkal jsem ho, prâvë když vycházel z knihkupectví na Karlovë námêstí. 2. Chodili jsme społu skoro hodinu. 3. Najednou mi fekl, že musi na schuzi, že prÿ už tam - tj. na druhé stranë mësta - na nëj nëkdo čekā, a odbëhl. 4. "Nikdy ho asi nepochopim", fekl jsem si. 1. (I) met him, just when (he) came out of the bookstore at Chari's square. 2. (We) walked together for almost an hour.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־-152

3. Suddenly (he) told me, that (he) must go to a meeting, and that someone is waiting for him there, that is on the other side of town, and (he) ran off. 4. "(I)'11 never understand him",(I) said to myself.

It has been shown in Chapter 3 that, with respect to the other two null-subject criteria, Czech also qualifies as a

canonical null-subject language. Recall that, in Czech, overt pronominal subjects cannot act as bound variables, as illustrated in (5) and (6) . Nor can they be interpreted as arbitrary in reference, as shown in (7).

(5) každy učitel, si mysli, že onVj má hloupého syna every teacher refi thinks that he has stupid son some specific־) 1every teacher thinks that he person) has a dumb son' (6) Petrj ještē stále miłuje tu ženu, kterou si Peter stili continue loves the women whom refi on.į/j pfed pēti Iety vzal. he ago 5 years took that other־) 1Peter still loves the woman whom he man) married 5 years ago' (7) v Nëmecku *oni ràdi pijou českė pivo in Germany they gladly drink Czech beer 'in Germany people like to drink Czech beer'

In section 4.3. following Faska (1964), a list of three environments was given in which Sorbian strongly prefers overt thematic subjects. They are repeated below under (8).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-153- (8) An overt pronominal subject is preferred in Sorbian a) if direct or indirect speech is being introduced b) if the matrix subject is a demonstrative c) if the matrix subject is a full NP

In Czech, on the other hand, overt pronouns are not at all

privileged in those environments, but instead are avoided as shown in examples (9-12). Environment (8c) directly violates one of the null-subject criteria. Example (9) below is thus related to the sentence type illustrated by (5) and (6) above. Sentences (10) and (11) contrast with (8a), and (12) contrasts with (8b).

(9) mi rodiče fekli, že pfijdou zitra my parents said that will-come tomorrow 1my parentSj said, that (they)^ will come tomorrow1

(10) fikal, že je unavenÿ saicĻ^ that aux^ tired^ 1(he) said that (he) is tired1 (11) "nebojím se ničeho", ïekl not-afraidltg refi nothing saičĻ^ I) don't fear anything", (he) said1)״1 (12) ta fekla, že je unavená ,״,tired ״that auxj ״*thatfan said 1that one said that (she) is tired'

These facts further support the claim that Czech is a canonical null-subject language, in contrast to Sorbian.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access However, elements which appear to be overt expletives do occur frequently in spoken Czech.6 Overt expletives are not found in stylistically neutral written texts; however, they can occur in text types which in general try to imitate the spoken language. Schalich (1970: 78-79), for example, shows how expletive pronouns are used as "expressive means of lan- диаде" in modern Czech comedies. There have been very few attempts thus far to account for this phenomenon. Trāvniček (1962) deals with it from a dia-

chronic perspective, while Koenitz (1988) and Hirschová (1984) try to provide an account within the framework of functional

sentence perspective. Both analyze the expletive as either

belonging to the thematic part of the sentence or marking the actual (very general) theme of the sentence. Franks (1995) and Lindseth (1994) are the only analyses within government binding, and Woolhiser (1994) provides a general survey of the syntactic and semantic characteristics of expletive-like pronouns in several Slavic languages.

Expletives in spoken Czech are optionally overt, so this is not an instantiation of the fourth language type charac- terized as impossible in the identification/licensing model. However, the linguistic facts seem to contradict previous predictions, e.g. Hyams*s, about the complete absence of overt

Townsend (1990) points out that there is a higher frequency of overt referential pronouns in the spoken lan- диаде. This is not too surprising, considering that more emotive/subjective factors enter into this type of speech.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-155- expletive subjects in a language with null referential pronouns. The examples below illustrate this phenomenon, which poses an interesting problem for the licensing/identification framework. In none of the sentences is the subject assigned a thematic role. The subject position thus seems to be occupied by an expletive pronoun. The form of the Czech expletive pronoun (v)ono is identical to the personal pronoun of the third person singular neuter.7

(13) a. vono je chladno 'it is cold' b. vono ještē präi 'it is still raining' c. vono se blÿskà 'it is lightening1

(14) a. vono se tam nepracuje it refi there not-works 'no work is being done there1 (cf. Es wird dort nicht gearbeitet) b. vono se tu tańcuje it refi there dances *dancing is going on there* (cf. Es wird dort getanzt)

One obvious phonological feature of Common Czech in contrast to Standard Czech is the use of a prothetic consonant [v] before initial о in most native Czech words, in- eluding the third person pronouns. Ono thus becomes vono. Hammer (1985) points out that in colloquial Czech this feature is used by virtually all speakers when it comes to pronouns and prepositions, but that its use in other words is more sensitive to the topic of the conversation, the type of words, the background of the speaker, etc. Townsend (1990:36) calls the prothetic shift о > vo "the most important SPC (Spoken Prague Czech) feature and certainly the most difficult to analyze", presumably because of the multitude of factors which influence its use. Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-156- (15) a. voao je možny, že tam nepfijede it is possible that there not-comes 'it is possible that he/she doesn't get there' b. vono je pfece nutnÿ, abysme ji to fekli it is yet necessary auxļpļsubj her that saicĻ, 'it is after all necessary that we tell her1 c. vono se zda, že to nebude lehkÿ it refi seems that this not-will-be easy 1it seems that this is not going to be easy'

An immediate question is whether or not these expletives are really subjects. Of course, the answer to this question

will depend on our definition of subject. Within Government Binding theory the notion of subject is defined configuration- ally as the NP which occupies the specifier position of IP.8 First of all, I assume that expletives have the categorial

status of NPs and are not some sort of particle or adverbial. They are thus eligible to occupy subject position. This assumption is based on the fact that their form coincides with that of the third person neuter personal pronoun, and, in the examples above, they seem to correspond to overt non-referential subject pronouns in non-null-subject languages (it in the English translation of the examples above). If the expletive is in subject position, then we would expect it to be compatible with material in С or [Spec,CP], as is the case for a normal

Under the split-INFL hypothesis, the subject position would be [Spec,AGRP]. However, this distinction has no bearing on the current discussion.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00062018

-157-

subject. The tree in (16) illustrates this for some English examples. (16) CP XP IP NP

V NP

whom, willj Peter tj invite t! I know that Peter will invite you does: lovehe you

For one thing, expletives in Czech seem to be more or less restricted to sentence initial position, which could be taken to indicate that they occupy the specifier of CP instead of [NP,IP] . If they are not structural subjects, then they should not cause any trouble per se for the licensing/identification framework, since there could still be an empty NP (pro) in [Spec, IP] (with the non-referential element occupying further up" in the sentence).10״ a different position

9 In English there is a restriction that a wh-word in [Spec,CP] and an overt complementizer under С cannot co-occur. In the literature this constraint is referred to as the "Doubly-filled COMP filter", dating back to a time when [Spec,CP] and С were not clearly distinguished. In several other languages, e.g. in Dutch and in Bavarian German, the sequence of a wh-word and a complementizer is possible. 10 For example, one subtype of German non-referential es as in (i) seems to be restricted to [Spec,CP], and is thus incompatible with wh-words such as warm (,when1), as shown in

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-158-

The examples below were judged by native speakers for their grammaticality. The complementizer protone in (17-19), když in (20) and the preposed verb nepracovalo in (21) occupy C, while the question words in (22-24) are in [Spec,CP].

(17) nemužu jit ven, protože vono tam prši not-can go outside because it there rains ' (I) can't go outside, because it is raining there1

(18) nikdo tu nebyl, protože vono se tu nobody there not-was because it refi there včera nepracovalo ,״״yesterday not-worked ,nobody was there because there was no work being done yesterday' (19) protože vono se zda, že to nebude lehkÿ because it refi seems that this not-will-be easy ,because it seems that this is not going to be easy' (20) ?? nemužeme tady bÿt, když vono se tu nepracuje not-canllfhere be if it refi here not-works ,we can11 be here, if there is no work being done*

(ii). (See also the discussion in section 5.3.3.) (i) es wurde wie immer viel gelacht und gesungen ,there was as usual a lot of singing and dancing going o n ' (ii) *wann wurde es wie immer viel gelacht und gesungen? Under this analysis sentence (i) actually has an empty NP (pro) in subject position, as illustrated in (iii). (iii) CP 'C י־ " - XP С '"“ IP 'I' ־־־־"""-Np I'"" *־ i V P es wurdej pro gelacht tj

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-159- (21) ?? nepracovalo se vono včera? refi it yesterday ^״not-worked 1wasn't there any work done yesterday* (22) *kde je vono chladno 1where is it cold1 (23) *kdy vono pršl/prši vono when it rains/rains it 1when is it raining1

(24) *proč vono se tu včera nepracovalo why it refi here yesterday not-worked^ 'why wasn't there any work done yesterday*

At first glance, the results of this investigation seem to

contradict each other. Native speakers were quite willing to accept complementizers (especially protone) preceding vono; however, constructions with material in [Spec,CP] were rejected. As I will discuss below in section 5.2.2., the marginality or unacceptability of sentences (20*24) may be due to pragmatic rather than syntactic reasons. I thus claim, based on the compatibility between vono and material in the complementizer position С in examples (17-19), that expletives in Czech are indeed in subject position.11

The Czech expletive is not only used in impersonal constructions, such as impersonal passives and weather verbs, but

Within the typology proposed by Franks (1990), spoken Czech is thus an example for [-OSP,+OSP] .

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-160- it can co-occur with overt referential NP subjects, as shown in (25-32).12

(25) vona ta myôlenka má nëco do sebe expl.fce this thoughtfen1 has something to itself ,this thought has something to it1 (26) vona se Marenka starà vo všechno expl.fanrefl Maienkafem takes-care of everything 1Mafenka takes care of everything* (27) vona se tu naskytla ta vosoba expl.fen refi there appeared that person^ *there appeared that person* (28) von mu otec všechno dovoli expl.n^c him father everything allows ,father allows him everything* (29) von ho nemine trést ^him neg-spare punishment ^״.expl *punishment won't spare him' (30) vono tarn bylo moc lidi ,there was^ many people^p )״».expl 'there were many people there* (31) vono se ti to lhani jednou vymstí this lying^once revenge *״expl.^refl. you ,this lying will come back to haunt you sometime' (32) von byl dfiv vchod z druhÿ strany expl.^was once entrance from other side 'there used to be an entrance from the other side'

12 Example (25) is constructed, example (26) is taken from Koenitz (1988), example (27) from Schalich (1970), example (28) and (29) from Siebenschein *s dictionary (1983), examples (30) and (31) from Hirschová (1984), example (32) from Townsend (1990) . All examples were tested on and accepted by native informants.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-161- There is no intonational break between the pronoun and the rest of the sentence, which indicates that this is not some type of left-dislocation.13 Moreover, the expletive in such constructions agrees with the associated referential NP in pronominal features. We thus have feminine vona in (25), (26), and (27), masculine von in (28), (29), and (32), and neuter vono in (30) and (31) . Example (30) is interesting with respect to agreement. It contains a quantified NP in the genitive case, so that default neuter singular agreement appears on the pronoun. The finite verb bylo is also in the neuter singular form. Both the expletive and the thematic NP appear in the

nominative case. Expletives in this construction are for the most part compatible not only with overt material in C, as in examples

(33) and (34), but also in Spec-CP, as shown in (35-37)u .

(33) myslím, že von« ta tvoję шуІІвпк* má пбсо do sebe 11 think that your thought has something to it1

13 In left-dislocation constructions, an NP is ad- joined to the left of IP and picked up again by a coindexed NP, as shown in the example below. binding principleSį, I cannot figure them* out Coindexation of left-dislocation constructions is not a violation of principle В or С (a pronoun has to be free in its governing category; an R-expression has to be free everywhere), since the binding theory applies to А -binding only and the antecedent in the left-dislocation construction occupies an A' position.

For glosses compare examples (25)-(28).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-162- (34) ... protože vona se Marenka starà vo všechno *because M. takes care of everything' (35) najednou vona se tu naskytla ta vosoba 'suddenly that person appeared there1 (36) kdy vona se tu naskytla ta voeoba 1when did that person appear here' (37) proč von mu otec vãechno dovoli 'why does dad allow him everything1

Based on these facts, I conclude that the expletive is in [Spec,XP] . The unexpected compatibility of the expletive with

material in [Spec,CP] in this construction type (in contrast to the vono prši/vono se tu nepracuje type of construction) will be discussed in 5.2.2. The next question which arises is which structural posi-

tion the referential nominative NPs in this "double subject" construction type occupy. According to Sportiche's (1988) VP- internal subject hypothesis, all arguments, including the subject, are projected in the VP. Thus there is another subject position, namely [Spec,VP] , which is assumed to be the D- structure position for all thematic subjects. I propose that the referential subject NPs in the examples under (25-37) actually remain in Spec-VP at S-structure. This is illustrated in (38) .15

15 At LF the referential NP moves up to Spec-IP to adjoin (or replace) the expletive, so that no violation of the principle of full interpretation occurs, according to which LF may only contain elements which contribute to the semantic interpretation.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access -163- (38) IP ז NP I

NP V 1 ו t V expl. ref. NP

The expletive and the referential NP associated with it presumably form some sort of CHAIN at S-structure. The members of this chain share pronominal features and case.

Alternatively, nominative case assignment to the expie- tive and the thematic pronouns could occur independently. As pointed out before, within recent analyses it has become standard to assume that what was formerly IP should be split into several specialized functional projections of V, such as AGRP, TP, NegP etc. Let us further assume with Roberts (1993: 27) that either AGR or T can assign nominative case as a matter of parametric choice. According to Roberts, Agr can assign case under spec-head agreement or government ECM; however, T can only assign case under government ECM.16 It is thus conceiv

ECM * Exceptional Case Marking. This case marking is "exceptional" because the case assigning head X assigns case to the specifier position of a maximal projection YP, Maximal projections normally constitute barriers for case assignment from outside.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-164- able that in some languages, including under this analysis

Czech, T assigns nominative case under government of the subject NP in Spec-VP. At the same time, AGR is able to assign nominative to its specifier, occupied by the expletive, under spec-head agreement.17 This alternative explanation of how nominative case assignment can be accounted for in the Czech double-subject construction is shown in (39) .

(39) ArgP

XP Agr1 Agr TP

T VP NP V expl. ref. NP

In principle, we could expect that a subject can use different Spec positions as landing sites, probably with cross-linguistic variation. Furthermore, it should be possible to independently lexicalize more than one specifier position within the (old) IP domain, as has recently been argued for Icelandic (Jonas & Bobaljik 1993) and Finnish (Holmberg & Nikanne 1994) . An alternative interpretation of the Czech construction type with expletive and associated thematic subject might then be that the referential subject occupies Spec-TP. Further research will have to show whether or not this is a viable option. It is far from incontravertible that all the different functional projections of V really exist. It would be convenient to prove that there is actually some "use" for the different specifier positions, such as Spec-TP or Spec-NegP.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־165־ 5.2*2. Origin and meaning of Czach expletives

In the previous section I argued that expletive pronouns in Czech occupy the structural subject position. We have thus found a counterexample to Hyams's (1986) prediction that languages which use referential pronouns for functional rather than grammatical reasons (i.e. null-subject languages) do not

have overt expletive pronouns. Czech is a canonical null- subject language which nevertheless uses lexical expletives.

However, it needs to be mentioned that most of the Czech constructions without expletives are not entirely synonymous with those which do have expletives. In fact, there seem to be three different types of expletive constructions.

The first group includes meteorological verbs and imper- sonai passives, as illustrated in (13) and (14) above. Here, an overt expletive is invariably associated with an emotive

involvement of the speaker, and always implies some sort of emotion such as surprise, joy, disappointment, etc. Above I analyzed vono type expletives as structural subjects, despite the fact that some speakers find the co-occurrence of such expletives with material in С awkward and most speakers even reject their co-occurrence with question words in CP. It seems to me that such incompatibilities can be attributed to pragmatic rather than syntactic considerations. A sentence such as Vono pršī expresses surprise on the part of the speaker about the fact that it is raining - something like the

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access English *Wow it's raining!1. It would not make much sense to turn this into a question *'When, wow, is it raining?'. The same native speakers consulted about the first type of construction also agreed that the expletive in the subject doubling construction in examples (25-32) does not necessarily contribute the same kind of emotive emphasis as the vono in the impersonal constructions.18 Instead, this type of construction seems to be used primarily to lower the style inten* tionally, depending on the speech situation, although a certain emotional involvement may also pertain. These findings are confirmed by Schalich (1970: 79), who writes about the role of this type of construction in the comedies of the Czech writer Karel Čapek (1890-1938) among others.

Es [the pronoun] ist hier ein rein aktualisie-rendes Element, das der Äußerung eine persönliche, leicht ־vertrauliche Note gibt und zugleich die Eindring lichkeit des Gesagten erhöht ... Im authentischen Dialog ist diese Diktion bisweilen durch ein ver- stárktes Kontaktbedürfnis des Sprechenden moti- viert.1v. . to

8 This seems to be more or less confirmed by other authors. Koenitz (1988) and Hirschová (1984) claim that vono type expletives in general have no emotive function. Howev- er, their analyses focus primarily on the subject doubling type Von mu otec všechno dovolî as in example (28) . 19 *It [the pronoun] is here purely an actualizing element which lends a personal, slightly intimate touch to the statement and at the same time increases the forcefulness of what has been said ... In authentic dialogue this diction is occasionally motivated by an increased desire for interaction on the part of the speaker.'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-167- Examples (29) and (31) above were generally interpreted as

having an additional emotional nuance of warning or threat, as

just you״ if the overt pronoun von/vono expressed a kind of wait" attitude on the part of the speaker. However, the main

function of the overt expletive appears to be to mark the discourse as more informal or friendly. It may be exactly the

lack of a clear emotive emphasis in most sentences of this

type which allows the expletive in this construction type to co-occur with wh-words, as illustrated in examples (36) and (37) above.20 The third type of construction was shown in the examples under (15). Here, the pronoun vono is associated (coindexed) with an extraposed finite clause. It seems to me that in these cases vono has actually more of the "pointing" function of a demonstrative and is thus not non-referential as such. In this type of construction vono can be replaced with the demonstrative to without any change in meaning.

(15*) vono/to je moinÿ, že tam nepfijede 1it is possible that he/she doesn't get there1

The use of pronouns in this position has been reported for other Slavic languages which otherwise do not allow overt expletives. Franks (1990) reports that in Russian the neuter demonstrative èto can be used in these contexts. Furthermore,

20 Most informants also rejected variants of the emotionally colored examples (29) and (31) when material other than the complementizer protoŽe preceded the expletive.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-168 - a demonstrative pronoun to is marginally possible in the same type of construction in Serbo-Croatian and Polish, The fact

that these languages use a demonstrative instead of the 3rd person pronoun in this construction indirectly supports my argument about the pointing function of Czech vono. The following examples are taken from Franks (1990: 19-21) .

(40) (èto) xoroäo, Čto Ivan poêël domoj IRu ] '(it) is good that I. went home*

(41) ?(to) je dobro. Sto Jovan ide kući [SC] '(it) is good that J. is going home1

(42) ?(to) jest dobrze, źe wszyscy poszli do domu [PO] 1(it) is good that everybody went home'

An optional use of the pronoun es can also be observed in German. As will be discussed in more detail in 5.3.3. German has obligatorily overt quasi-argumentai subjects (weil es/*pro regnet) and obligatorily null non-argumentai subjects (weil рго/*ев hier getanzt wird) . However, in the context of an extraposed finite clause this complementary distribution seems to break down. Both the variants with and without es are possible, as shown in the sentences below.

(43) ... weil möglich ist, daß er morgen kommt ... weil es möglich ist, daß er morgen kommt ...*because (it) is possible that he comes tomorrow' (44) ... weil schön ist, daß du morgen kommst ... weil es schön ist, daß du morgen kommst ...*because (it) is nice that you will come tomorrow'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-169- I thus argue that the occurrence of vono in the construction type with an extraposed finite clause represents a different phenomenon which is not immediately related to the possibility of overt expletives in a language. However, taking into account the fact that the acceptability of the constructions in Serbo-Croatian and Polish is at best marginal, the occurrence of "real" expletive subjects in Czech may well affect the use of vono in these constructions.21

The functions of the expletive in the first and second type of construction are related in that they express the involvement/attitude of the speaker. This sheds doubt on Hyams's (1986) position that non-referential elements (i.e.

21 Franks (1990) treats the pronouns in constructions with extraposed finite clauses as *■true" expletives. Recall that he derives the null-subject properties of a language from its nominative case assignment properties (Nominative case must be realized? [OESP]; Nominative case cannot be assigned at LF? [OSP]) Null expletives are generally irrele- visibility of arguments. If a language does־vant for the LF not obligatorily assign nominative case, expletives in that language do not get case and thus cannot be lexicalized. However, Franks predicts that a certain type of overt expie- tive should be possible even in [-OSP,-OESP] languages. Clausal and prepositional complements also need case for their theta-roles to be visible at LF. While infinitīvais and PPs have intrinsic case, finite complement clauses must form a chain with the subject position which receives nominative Case. The expletive in the theta-marked chain with a finite clause will thus receive case as well. At this point the expletive pronoun is like any referential pronoun which must have case for LF-visibility but does not need to be lexicalized due to the [-OSP] setting. It may, however, be lexicalized. Although not explicitly mentioned by Franks (1990), the difference in the acceptability of such elements in Czech on the one hand, and Polish and Serbo-Croatian on the other, might be attributed to the fact that Czech is [-OSP,+OESP] and thus allows optional expletives in general.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access expletives) cannot be used for functional/pragmatic purposes. Instead* I claim that overt expletives in Czech should be

subsumed under the generally possible occurrence of overt subject pronouns in marked discourse. Just as with referential subject pronouns, the use of an overt expletive is evoked by functional considerations. One might interpret lexicalized subject pronouns in Spec-IP (i.e. [Spec.AGRP]) in null-subject languages as an overt realization of the pronominal features of Agr to specially mark the discourse. The occurrence of overt expletives in Czech can thus be shown to be compatible with the avoid pronoun principle.

Recall that the use of overt referential pronouns did not contradict the principle, as long as they were used in emphatic contexts. Similarly, by the avoid pronoun principle, a lan-

диаде that licenses null subjects will never use overt expie- tives unless they convey some special pragmatic meaning. Within the scope of this work I will not attempt to analyze the different shades and nuances of this pragmatic meaning. In the following I will use 1'emphatic expletives" as a general term. Spoken Czech is special in having such expletives. The question as to why expletives in Czech, in contrast to other null-subject languages such as Italian or Spanish, are not really meaningless will be discussed below. The table below is an "updated" description of possible language types within the licensing/identification framework (cf. also Table 1) . It should be emphasized that the following

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-171- combinations are not parameter settings, but instead descriptions of possible language types derived from the settings for מ.[the actual parameters [pro licensed] and [pro identified

Table 3: Possible Language Types (III) i: canonical pro-drop A [±T,-E] e.g. Italian В [±T,±E] emphatic expletives, e.g. Czech ii: mixed languages С [+T,-E] e.g. Russian, German23

D [+T,±E] e.g. ?? iii: canonical non-pro-drop

E [+T,+E] e.g. English

Types A and В are canonical null-subject languages in the

sense that pro is licensed and identified. Both type С and D

belong to the mixed type, i.e. pro is licensed but not identi-

-to mean obligatorily overt, 1-,obligato * +״ I take 22 rily null and ' ± ' optionally overt. 'T1 stands for *thematic subjects1, *E* for 'expletive subjects'. 23 Although both German and Russian are mixed pro-drop languages, they are still different with respect to their null-subject properties. In German, quasi-argumentai subjects have to be lexicalized, while in standard Russian they must be pro, as illustrated in the contrast between (i) and (ii) . (i) ...weil es/*pro hier kalt ist 'because it is cold here* (ii) ...potomu čto *ono/pro zdes' xolodno As mentioned in the introduction, quasi-argumentai subjects share properties of both referential subjects and expletive subjects, so that such a cross-linguistic 1*vacillation11 with respect to overtness or covertness should not be too surprising.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access fied. In contrast to type C, type D would be a mixed pro-drop language with emphatic expletives. As of yet, this type has not been attested, although it is theoretically possible.24 The discussion thus far has shown that the Czech data can actually be accounted for by the licensing/identification framework and that the occurrence of optionally overt expletives in a language which licenses pro does not necessarily constitute a violation of Chomsky's avoid pronoun principle.

At this point it seems appropriate to point out some simi- larities and differences between the predictions made by the (modified) licensing/identification framework summarized in Table 3 and Franks's case-assignment based approach discussed earlier in this chapter (cf. Table 2). Both systems predict the existence of the Czech type, i.e. optionally overt expletives in a canonical null-subject language. According to an analysis à la Franks, spoken Czech has a different parameter setting than a null-subject language which does not allow overt expletives. Within a licensing/identification approach, there is no differ- enee in the parameter settings, and spoken Czech is unusual in that it has emphatic expletives. So far, only very few null- subject languages with overt expletives have been documented. For Franks, the question is why the setting [+OESP.-OSP] seems

2 Woolhiser (1994) mentions the occurrence of expie- tive pronouns in Ukrainian, Belorussian and Russian dialects. In Chapter 3 I analyzed Russian as a mixed pro-drop language. More research is necessary to determine whether or not these dialects may in fact be examples of language type D.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access -173- to be so rare.25 For my analysis, the exceptional status of Czech is justified by the fact that languages normally do not develop emphatic expletives. Franks*s system, based on two binary parameters, predicts the existence of only four Ian-

диаде types (cf. Table 2). Franks (1990: 14) mentions explicitly that the configuration which is predicted to be impose!- ble under his approach is a language with obligatorily overt thematic subjects and optionally overt expletives, which corresponds to type D in Table 3. The existence of mixed pro- drop languages with overt expletives (which have not been attested yet) would thus disprove the case-assignment analysis. For Franks, the justification for overt expletives in a ־־language is a setting of t+OESP] , and it depends on the set ting for the [OSP] parameter whether the expletives are obiig-

atorily ([+OSP]) or optionally ([-OSP]) overt. Optionally overt expletives derive from the setting [-OSP.+OESP] . On the other hand, for a language to have obligatorily overt thematic pronouns, the setting [+OSP] (nominative assignment cannot take

place at LF) is assumed, which leads to a contradiction. Expletives by definition lack any referential properties. How is it then possible that such an element can carry a prag- matically marked reading? The following discussion and exam-

25 Unless, of course, we assume that even in this language type the overtness of expletives is conditioned by functional considerations. It could then be possible that there are many languages with the setting [-OSP.+OESP] which simply could not be differentiated from [-OSP,-OESP] languages due to their lack of emphatic expletives.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access pies are based on TrávníCek (1962: 16-30), who gives a plausible historical account of the use of expletive pronouns in Czech. He argues that the emotive/emphatic meaning developed out of the upozoïiîovacî citoslovce (1attention interjection1) опо/ало, which corresponds to modern Czech hie or aj (1look'; 1behold1). In older Biblical texts ono/ano was originally used as one of the possible translations of Latin ecce. This element had sentential value on its own and was therefore positioned *outside" the CP of the following clause. The

Czech example in (45) illustrates this.

(4 5) James: 5:7 ano, tēžaf Čakā drahého óvoce (evangeliáf Olomouckÿ, 14th century; cited by Trávniöek (1962: 20)) ecce agricola expectat pretiosum fructum

•see, the farmer is waiting for the precious fruit*

aj , orāč očekāvā drahého uSitku zemského (Kralická Bible, 1613 edition)

Later, this element lost its original function and meaning. The development of its various emotional nuances was accompanied by the loss of its sentential value and incorpora- tion into the sentence structure proper of the following sentence. The original intonational break between the interjection and the next sentence was lost.26 According to Trávníőek,

26 More evidence for the change in the status of ano/ono comes from the placement of clitics. Recall that Czech has second position clitics. Trāvniček (1962: 21) writes:

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-175- agreeing forms, which in our terms reflect a movement "down the syntactic tree" into subject position, started to develop in the late 15th century. Examples (47) and (48) date back to the first half of the 16th century. There are also examples of non-agreeing ano/ono during the transitionary period, as illustrated in example (46) which dates back to the second half of the 14th century.27

(46) ano však i ta sasa nebesą vyznávají di vy i svatost jeho ,even the heavens confess his miracles and sainthood1

(47) a oni,! Pol&cip! volili jsű sobë za kní2e mladence j ednoho 1and the Poles elected a young man as their ruler'

Kdy2 citoslovce ztratilo vëtnou platnost, vyslovovalo se splÿvavë s následující vëtou a citilo se za její člen, začaly se klást pfiklonky hned za në. ,When the interjection lost its sentential value, merged with the following sentence and was felt as its member, the clitics began to attach immediately following it [the original interjection, ML].1 Unfortunately, no examples of clitics and "independent" ano/ono are available at this point, so that example (i) is a theoretical reconstruction based on (ii) which is given by TrávníCek (se is a reflexive clitic). (i) ano, jiS se setmievá ,behold, it*s already getting dark1 (ii) ano se jiS setmievá 27 These examples are cited from Trávníãek. The difference between the ano in (46) and (45) is somewhat difficult to prove. Trávníõek bases his judgement on the lack of the comma in the former. A comma seems to be consistently present in the older examples where ano is still an independent citoslovečnā vēta (interjectional sentence). Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access -176- (48) onipi všichni öertip, vyšli sű z pekla pro tu duši 'all devils came out of hell for this soul'

In modern Czech agreement is more or less obligatory, unless of course there is no subject with which to agree, which produces the vono prŠ£/vono se tu tańcuje type.28 How- ever, some modern dialects may have preserved an intermediate stage of this development with a non-agreeing form of the expletive. TrávníCek cites, among others, the Czecho-Moravian dialects as an example, as shown in (49) .

(49) vono Casto lidi vidëli песо ехр1.шч often people^ saw something 'people often saw something'

TrávníCek (1962: 25) calls the pronoun which evolved out of the interjection naladové on ('moody it') and maintains that it expresses the ci tövé zaujetî ('emotive bias') of the speaker whenever it is used. While this is true for the vono prši/vono se pracuje type, it contradicts my findings somewhat as far as the double subject type is concerned. Recall that

" I can only speculate about why this type of construction is not part of the modern spisovná čeština. One important source for the codification of the literary lan- диаде in the early 19th century was the Kralice Bible from 1593, which was based on the Moravian dialect. (Kralicy is a town in Moravia). I did not find a single example of subject doubling constructions in the Gospel according to St. Matthew as translated in the Kralice Bible, which leads me to con- elude that this type of construction in general was not typical for this Bible translation. Recall that the phenomenon of overt expletives was discussed as a feature of the spoken language (obeenâ Čeština or bēžnē mluvená ČeŠtina) of Bohe- mia.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-177- most speakers agreed that the expletives in examples (25-32) above do not necessarily contribute an emotive emphasis. However, it is possible that the nâladovy odstin (1emotional nuance1) has further changed in the thirty years since Trāvniček wrote his book.

As for the use of vono in constructions with extraposed finite clauses, I speculate that this function dates back directly to the original common Slavic demonstrative pronoun опъ, ona, оло. Common Slavic had no 3rd person pronouns, and it was this demonstrative which developed into 3rd person

pronouns in most Slavic languages. Czech is unusual in that

it seems to have had an additional source for the 3rd person pronoun, namely the ирогогйоva ci citoslovce mentioned above.29

5.3. Upper Sorbian In Chapter 4 I argued that Upper Sorbian does not behave like a canonical null-subject language with respect to the occurrence of referential null pronouns. Sorbian thus produces a serious challenge for a licensing/identification approach to null-subject phenomena. In this section I will investigate whether or not the licensing/identification theory, especially

-Trávínãek seems to imply that all 3rd person pro מ nouns are derived from the original interjection ono/ano (which in turn is most likely derived from the Common Slavic demonstrative pronoun ono) . He maintains that the double subject constructions occurred at a time before the appearance of normal 3rd person pronouns and basically paved the way for their introduction into the language.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access the morphological uniformity hypothesis, makes correct predictions about non-referential subject pronouns in Sorbian.

5.3•1. Syntactic statue of overt expletives The Sorbian verbal paradigm fulfills the morphological uniformity criterion, and thus we would expect pro to be licensed. In keeping with Chomsky's avoid pronoun principle and based on the discussion in section 5.2.2. we would then expect that the language should not use overt expletive subjects unless they are able to fulfill some sort of pragmatic function, as we have seen in spoken Czech. Overt non-referential pronouns are widely used in spoken Sorbian, although they are more or less absent from the literary language. Schuster-Šewc (1974) was, to my knowledge, the first linguist to attempt to give a systematic account of this phenomenon. A discussion of several aspects of his article will be included in this section.

The different forms of the expletive are wono and wone. These forms are distributed according to dialectal areas. Wono is typical for Lower Sorbian, the transitionary dialects (Muskau, Schleife, Spreewitz, Hoyerswerda), and the Northern dialects of Upper Sorbian (Nochten, Reichwalde, Klitten, Kreba), and wone for most other Upper Sorbian dialects.30 In

30 Schuster-Šewc (1974) points out that the form wone is a secondary form derived from wono, probably under the influence of the distinction between to and te for the nominal and neuter attributive use of the demonstrative pronoun.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-179- the South-Eastern tip of Lusatia the form wón, which is homon-

ymous with the 3rd person masculine form, is used as a non- referential pronoun instead of wone. This phenomenon seems to be strictly limited to a very small area, and speakers from other dialect zones reject it outright. The examples (50-57) below were collected in the "Kemgebiet" .

(50) Jana je prajiła, zo chce wona rady kupać hić. 1J. said, that she would like to go swimmming.* Won• pak je hišče pfezyma. it however is still too-cold *However it is still too cold* (51) won• je Škoda, zo wjeőor njepfińdźeS it is pity that evening not-will-come^ *it is too bad that you won't come in the evening*

(52) won• njeje rjenje, zo je pfeco hišče tak zyma it not-is nice that is always still so cold

*It is not nice that it is still so cold' (53) poprawom smy chcyli dźensa pućować, ale won• at-first auxip, wanted today hike but it so nažel deščuje refi unfortunately rains *originally we wanted to go for a walk today, but unfortunately it is raining* (54) won• je móSno, zo wón hišče pfińdźe it is possible that he still will-come 'it is possible that he will still come'

(i) to ma chwile 1this had time' (ii) te blido *this table' Wone first appears in Swëtlik's 1690 Bible translation. Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־180־ (55) My hakle jaxa pozdźe do wjeski pfińdźechmy, we only very late to village came a wone bēše hižo ćma. Wone njebēše nam cyle der je. and it was already dark it not-was to-us quite good 'We came into the village very late and it was already dark. It was a little scary.' (56) dźensa so prózdniny započnu; wone je tež wulki čas today refi vacation start it is also high time 'today the break starts; it is also high time' (57) wśitcy bëchu wjeseli; wone je so wjele smjeło, all were happy it was refi much laughed rejwało a spëwalo danced and sung *everybody was happy; there was a lot of laughing, dancing, and singing'

Consultants with some degree of formal language training, such as teachers, ministers, and clerical workers, claimed not to use this sort of construction, but did not deny familiarity with the phenomenon. Prof. Helmut Faßke of the sorbisches Institut analyzed native speaker recordings of spontaneous speech from a large variety of dialects in preparation for the syntax volume of the Sorbischer Sprachatlas (Volume XV to

appear) and kindly shared some of his data with me. The following examples are taken from these recordings.31

(58) wono jo se tak wjele Sło, žo äycko spódy wódy stoj *it rained so much that everything was under water'

־Some dialect forms cited do not exist in the stan 31 dard language. For further examples of this type see also Schuster-Šewc (1974).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-181- (59) wono było žon jare pozdźe 'it was already very late1

(60) pon bu wone wordowało mudrowane a powëdane ,then there was a lot of smart talking' (61) wone je dźensa jara mało widźeć ,there is very little to be seen today'

(62) ale wone bëäe nuzne, zo so ta rëka zahaći 1but it was necessary to dam up that river'

(63) wone dyrbješe Cisće znowa započane wordować 1it had to be started all over again1 (64) ale wone tež skoro zas tak je kaž pojėda 'but it is also almost as they say'

(65) wón bë to pon tak daloko pfidło, zo .... 1 it went so far that *..'

Based on the discussion in section 5.2.1., the pronouns in examples (51), (52), (54), and (62) should perhaps not be interpreted as "pure" non-referential pronouns, and instead be seen as being similar to the Czech vono co-occurring with extraposed finite clauses. In the following I will concen- träte on the other environments in which the expletive-like elements seem to occur.

The use of non-referential pronouns is optional. In clear contrast to Czech, no emotive/emphatic component is associated with the use of these expletives. Thus, if they

turned out to be "real" subjects this would be in conflict with the theoretical predictions outlined above. In particular, we would have found a language which lexicalizes non-

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־182־ emphatic expletive subjects, although pro is licensed. However, in contrast to Czech, there seems to be a syntactic incompatibility between expletives and wh-phrases or complementizers. This hypothesis is supported by evidence from various sources. Faßke's corpus of data, for example, provides over 50 examples of overt expletives in many different dialect areas. However, only a single case can be found where wone was used after a complementizer. I also asked informants to translate sentences containing overt expletives

from German into Sorbian. There was not a single instance in which speakers used an overt expletive after a complementizer

h-word, even if there was an es in the German original.32׳or w The sentences below, produced by an informant who tended to use overt expletives very frequently, illustrate this.

(66) wone pak je hiŠče pfezyma it however is still too-cold ,it is still too cold1 (es ist aber noch zu kalt)

(67) mi je žel, zo pfeco hišče tak zyma je me is sorry that always still so cold is ,I'm sorry that it is still so cold' (mir tut es leid, daß ев immer noch so kalt ist)

All Sorbian speakers rejected constructed examples where wone/wono was used after a wh-word or complementizer. This is shown in examples (68) and (69) .

For further comparison between Sorbian and German see section 5.3.3.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

183-־ (68) hdy (*wone) je zyma 1when is it cold*

(69) džēči su zrudne, dokelž (*wone) hižo taje 1the kids are sad because it is already thawing1

In contrast to Czech, in Sorbian there is a very clear syntactic incompatibility between overt expletives and overt material in the head or specifier of CP. In view of the lack of any sort of emotive component inherent to these pronouns, it is not an option to attribute these distributional restrictions to pragmatic reasons. This leads me to conclude that Sorbian expletives are positioned outside [Spec,IP], most likely in [Spec,CP]. Presumably, pro is occupying [Spec.IP] referential wone/wono is present, giving־ even if an overt, non the structure (70) below.

(70) CP XP^

IP NP VP ז I 1 wone pro taje it thaws

To continue this line of reasoning it is useful to look at the two different types of conjunctions. Subordinating conjunctions such as dokelž (,because1) in example (71) reside in the complementizer position С and take a subordinate clause (IP) as complement. The expletive wone, occupying [Spec,CP]

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-184- lies outside this IP, and a complementizer placed in front of wone would thus lead to branch crossing and therefore violate X-bar structure.33 Coordinating conjunctions such as ale

(1but1), on the other hand, conjoin two elements of identical category, such as two CPs and should therefore not inhibit the occurrence of the expletive wone. This expectation is borne out by the next pair of sentences, produced by the same informant (sentence (53) is repeated as (72)).

(71) wón njeje kopańcu hrać Soł , dokelž he neg-aux3ií soccer play went because je so deSćowało aux3}| refi rained 1he didn*t go to play soccer, because it was raining' (72) poprawom smy chcyli dźensa pućować hić, ale wone so nažel deSćuje ,originally we wanted to go for a walk, but it is unfortunately raining'

A curious type of construction was discovered in a dialect text from the Mješicy region, cited by Faßke and Michał к (1989: 41). Example (73) is an arbitrary plural construction

The sequence *wone dokelž1, although theoretically conceivable in terms of the tree structure, is not possible. Also compare the ungrammaticality of the German sentences below : (i) *weil es getanzt wird ,because it is being danced* (ii) *es weil getanzt wird

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-185- with a 3rd person singular expletive in initial position.34

This indirectly supports my hypothesis that the overt expie- tive in Upper Sorbian is not in [Spec,IP] ( [Spec,AGRP] ), since, otherwise, we would be faced with a very unusual situation of

non-agreement between the singular pronominal subject wón and the plural verb su pójdali.

(73) wón su jow wele pójdali, zo expl.jg auXp, here much saicĻ, that ♦ 1one used to say here often that1

An analogous example was found among the tapes analyzed by Faßke, cited here as (74).

(74) wón to takle činjachu expl.4 that so didp, *it used to be done that way1

5.3.2. Historical perspective It follows from Seiler's (1830: 113-114) description of the phenomenon that the situation in the early 19th century seems to have been more or less the same as in today's spoken lan- диаде.

Das unbestimmte Pronomen und unbekannte Subjekt Es wird durch wono oder wone gegeben, welches dann das erste Wort zu sein pflegt: wono rima, es donnert; wono s deščom a se snëhom dźe, es regnet und schneit zusammen; wone zwoni, es läutet; wone so wujasnjuje.

M Mješicy (Meschwitz) is in the Sorbian speaking territory which uses the form wón as an expletive pronoun (see 5.3.1.).

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00062018

-186- es hellt sich auf. Da aber schon die dritte Person singul. neutr. des Zeitworts das Es bezeichnet, so ־wird in Sätzen, wo Konjunktionen oder andere Rede theile voranstehen können, wono weggelassen: hdyž rima, wenn es donnert.33

Another indication that these elements were widely used in the literary language is Smoler*s (1841) Gesprächsbuch. J. A. Smoler was one of the leading figures of the serbske narodne wozrodźenje in the 19th century and was particularly concerned about the creation of a modern, unified Sorbian literary language. The following are some examples found in Smoler (1841: 12) (examples are given in modern spelling).

(75) wone so błyska es blitzt wone so počina hić es fängt an zu regnen wone (so) próSi es staubt wone (so) kapje es tröpfelt wone so lije es gießt

Compare this to the Gesprächsteil in Krai's (1895) grammar

(76) chłódno je es ist kühl hižo pfekapuje es fängt an zu regnen je suche, mokre es ist trocken, naß

The indeterminate pronoun and unknown subject Es' מ (1it1) is given as wono or wone, which is then the first word in the sentence: wono rima, es donnert (,it is thundering*); won s deščom a se snëhom dźe, es regnet und schneit zusammen ('it is raining and snowing at the same time*); wone zwoni, es läutet ('it is ringing1); wone so wujasnuje, es hellt sich auf (*it is getting lighter*). However, since the 3rd person of the verb already shows the Es ('it'), wono is omitted in sentences with preceding conjunctions and other parts of speech: hdyž rima, wenn es donnert ('when it is thundering*).'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-187-

je jara ćopło es ist sehr warm pfeco hiščo mjerznje es friert immer noch

The use of wone thus seems to have disappeared from the literary language in the second half of the 19th century. It is

virtually non-existent in today's written language. It seems fair to assume that the puristic movement associated with the serbske narodne wozrodżenje artificially caused this sudden disappearance of overt non-referential pronouns from the literary language. Schuster-Šewc comes to a very similar

conclusion when he writes (1974: 343):

Das Meiden des fiktiven Subjekts wono(e) in der neueren os. Schriftsprache dürfte zweifellos mit der seit Mitte des vergangenen Jahrhunderts verstärkt auftretenden Tendenz verbunden sein, die Schriftsprache stärker von der Volkssprache abzusetzen und damit auch fremde (d.i. deutsche) sprachliche Einflüsse zurückzudrängen.36

According to Schuster-Šewc, overt expletive pronouns appear after 1850 sporadically in those literary works which intentionally reflect the Volkssprache. As a typical example he cites the writer A. Cernÿ, whose legends about Mythiske bytosče lužiskich Serbów (1Mythical Beings of the Lusatian Sorbs') were published between 1890-95. This example may not be the most appropriate; Cernÿ was Czech by birth and might

Ä 'The avoidance of the fictitious subject wono(e) in the newer Upper Sorbian written language is undoubtedly connected to the strong tendency, starting in the middle of the last century, to separate the written language further from the vernacular language and thereby to push back alien (i.e. German) linguistic influences.'

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access have introduced some linguistic "interference" with respect to the use of overt expletive pronouns. However, this should not affect the general validity of Schuster-Śewc1s statement. Schuster-Śewc (1974) points out that in some Old Sorbian manuscripts there are examples of expletives which co-occur with referential subjects in the same sentence. Often they agree in pronominal features with the referential subject, making this phenomenon very reminiscent of the spoken Czech construction discussed above. The most typical text for this is Swëtlik's (1690) translation of parts of the New Testament, from which the following examples are taken.

(77) Matthew 8.26

. . . , ha wonafem so sta jena wulka ćichotafem and/expl./refi./happened/one/great/calmness (78) Matthew 8.32 (cited from Michałk (1972: 93) ha hlejćo, wona*™ z huj om woteńdźe ta су I a SrjÓdafo,, pfez nahły spad do toho mor ja nut? and/behold/expl./with/quickness/went-away/ that/whole/herd/over/steep/fali/to/the/sea/inside

It is interesting to compare Swëtlik's 1690 translation of various Biblical passages with the 1750 version of the same passages. The latter, according to Soł ta (1921: 33), should be considered as an edited Swëtlik translation rather than a completely new one. Surprisingly, the constructions with agreeing expletive-like elements seem to have virtually

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-189-

disappeared in the 1750 pericopes. The examples below provide some direct comparisons.

(79) a. Matthew 8:13 (Swëtlik 1690)

: k tomu hawtmanej ^״ha w ó n ^ rjekny JSzue wustrowjeny wo ^״ha w ó n ^ bë tón hólc ... tej samej štundže. and/expl./said/Jesus/to/the/centurion/ and/expl./was/that/boy/healed/in/same/hour b. Matthew 8:13 (1750 pericopes)

ha JSzuø rjekny tomu hauptmanej: ... ha tón hólc bó wustrowjeny wo tejsamoj štundže. (80) a. Matthew 8:8 (1690)

ha wón^ budźo bóć mój wotroök^ wustrojeny. and/expl./will/be/ту/servant/healed b. Matthew 8:8 (1750) ha budźe mój sluŽownik wustrowjeny. (81) a. Matthew 8:25 (1690) Ha wonipj pfistupichu joho huSownikip! к njomu. and/they/stepped-to/his/disciples/to/him b. Matthew 8:25 (1750) Ha ći joho wućownicy stupichu к njomu. (82) a. Matthew 8:3 (1690) ^״Ha wónnułC bë hnydy tón joho wusad wuSwarnjeny. and/expl./was/immediately/this/his/leprosy/healed b. Matthew 8:3 (1750)

Ha hnydy bó ton joho wusad wuSwarnjeny.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-190-

Since this difference between the two editions could not yet be due to puristic endeavors, it confirms my hypothesis that the construction with agreeing expletives was never really Instead, it was more or less part ת.typical for Old Sorbian of Swëtlik 1s personal idiolect. Perhaps it was felt to be strange and thus removed by the editors of the 1750 version. One might speculate that Swëtlik was particularly influenced by Czech constructions with agreeing expletives since he spent about ten years in Bohemia. However, a comparison with the analogous passages from the Wenceslaus Bible (1677) shows that this special feature of Swëtlik's language could not be due to direct influence from the Czech Bible. As already mentioned above, there are also no instances of such expletives in the older Kralická Bible (1613) . It is thus more likely that Swëtlik was influenced by spoken Czech.

5.3 * 3 . Upper Sorbian and German When considering the possible origin of non-referential pronouns in Upper Sorbian, one may consider German influence as a possible source.

(83) es regnet; es ist kalt; es ist dunkel

However, as first illustrated by Schuster-Šewc (1974), the

Also recall that there was only a very insignifi- cant decline in the number of overt referential pronouns in the 1750 version compared to Swëtlik (1690) . For details see 4.4.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־191־ syntactic distribution of German es and Sorbian wone/wono is notably different. First of all, the occurrence of overt non- referential pronouns in Upper Sorbian is always optional, even in those dialects which strongly prefer the version with the expletive. In German, certain non-referential pronouns, -weather״ namely those which are subjects of the so-called verbs", are obligatorily overt, independent of their position in the sentence.

(84) a) «8 regnet morgen 'it is raining' b) wann regnet вв 1when is it raining1 c) weil es regnet 'because it is raining1

As explained above, in Sorbian, the expletive is barred from occurring in the syntactic positions in (84b) and (84c) because of the structure posited in (70) . In German, the quasi- argumentai es which is associated with meteorological verbs word in (Spec,CP] or־can (and in fact must) co-occur with a wh a complementizer in position C. These distributional proper-

ties indicate that this type of German es is indeed the struc- turai subject of the sentence.38

58 Es in these clauses is base-generated in [NP.IP], but may move to a different position at surface structure. The following structures assume that subordinate clauses represent the underlying word order and that root clauses are derived by obligatory movement of the finite verb into C. CP CP 'C ־־ ־־־־ C' N P־"־'"' 'NP ip С IP י -'*״־•י — - С 'I "־־I' NP '׳־־ ־"NP weil es regnet es! regnetk t i t k Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 192-־ Michałk (1965a, 1977) cites some interesting instances of "Sorbisms" in German. Below are examples where Sorbian native speakers omit the (obligatory) German es, cited from Michałk (1965: 713) (# indicates the place where the es is missing).

(85) und wenn [#] dann soweit ist, dann 'and if [it] then is time, then1 (86) sehr kalt ist [#] bei uns nicht mehr. '[it] is not very cold anymore here1 (87) tagsüber taut [#] noch *during the day [it] is still thawing1

There is another type of German es which seems not to be a structural subject but resides rather in the same structural position as the Upper Sorbian non-referential wone, i.e. in [Spec,CP]. This is illustrated in the impersonal sentences below.39

(88) а. ев wurde viel getanzt und gesungen 'there was a lot of dancing and singing1 b ..... weil (*es) viel getanzt und gesungen wurde c. wann wurde (*es) viel getanzt und gesungen

Examples (88b) and (88c) are ungrammatical for exactly the

CP XP C' C* IP NP ‘ I1 wann regnetį es tį 39 Cf. also fn. 10.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-193- same reasons that Upper Sorbian wone and co-occurring subordinating conjunctions or wh-words are ungrammatical. This type

of es is not a structural subject and therefore is not generated in [Spec, IP]. Instead, it seems to be inserted into [Spec,CP] due to language specific requirements that this position has to be lexically filled in German declarative root clauses. As illustrated earlier, sentence (88a) actually has an empty NP (pro) in subject position [NP,IP], which identi- fies German as a mixed pro-drop language. Referential and quasi-argumental subjects have to be overt; non-argumental subjects are obligatorily pro. The German '1non-subject11 es and Upper Sorbian non-referential wone seem to serve the same purpose. In terms of fune- tional sentence perspective they occupy the theme position, presumably to allow the core of the message to be placed toward the end of the sentence. Some Upper Sorbian speakers who reject the use of overt expletives consistently placed wonka ('outside') in sentence initial position, apparently to achieve the same results. The following examples were elicited in German.

(89) wonka so deščuje - es regnet 1outside it rains' (90) wonka je zyma - es ist kalt ,outside it's cold1

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-194- 5.4. Summary In this chapter I analyzed the occurrence of expletive-like overt pronouns in Czech and Sorbian, which at first glance

appear to pose problems for a licensing/identification system for null subjects. According to standard views of null-subject

properties, the optional occurrence of overt non-referential pronouns is unexpected in any language where pro is licensed. I showed, first, that spoken Czech, despite its status as a canonical null-subject language, allows overt expletive subjects. However, because the use of Czech expletives is based

on functional considerations, I argued that this phenomenon does not really contradict theoretical predictions made by the licensing/identification framework in conjunction with the avoid pronoun principle. Sentences with and without overt expletives are usually not totally synonymous, and it thus seems that Czech expletives can convey some special pragmatic meaning. Therefore they should be subsumed under the generally possible occurrence of overt subject pronouns in marked context. I then followed Trávníöek's argument about the possible derivation of such "functionally charged" expletives

from a sentential interjection with the function of "drawing somebodyfs attention to something".

In contrast to Czech, no emotive component is associated with the use of expletives in Sorbian. However, the syntactic incompatibility of expletives with wh-phrases and complementizers indicates that expletives are not structural subjects.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-195- Instead I argued that these elements occupy [Spec,CP], presumably with pro in the subject position, which produces the structure [CP wone [IPpro taje]] ('it's thawing'). The

positioning of Sorbian wone is analogous to and perhaps under the influence of German non-argumental es (which crucially does not include the subject of weather-verbs) which is inserted in Verb-second contexts and only legitimate in that context. In contrast to Sorbian, German has another type of non-referential es, the so-called quasi-argumentai es, which is a real subject and must always be overt. This type of es is the subject of meteorological verbs in German. It follows then that in Sorbian, there is no difference in the position of

wone in the following examples, while in German there is a different position of es.

(91) a. wone taje 'it is thawing' (a, wone [fppro taje]]

wone je so wjele smjało 1there was a lot of laughing going on* [qp wone je so [IPpro wjele smjało]] (92) a. es taut [jp es taut] ]

b. es wurde viel gelacht [cp es wurde [1Ppro viel gelacht] ]

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00062018

-196- 6. CONCLUSION

In this work I discussed data from Russian, Sorbian, and Czech which appear to challenge licensing/identification systems such as the one suggested by Jaeggli/Safir (1989). The application of a set of three null-subject diagnostics to Slavic languages in Chapter 3 suggested that most West and South Slavic languages are canonical pro-drop languages, while Russian, taken to be representative of East Slavic, is not. After suggesting an intuitive explanation for this split, I set out to formalize the difference between these two groups in terms of the inconsistent realization of the functional .in Russian ״projection AGRpç

In Chapter 4 I showed that Sorbian does not possess all the properties of a canonical null-subject language, despite claims to the contrary by language purists. To explain and illustrate this puristic error I cited numerous examples from the written language (drawn both from early and contemporary texts) and the spoken languages. This was followed by a crit- ical discussion of three possible answers to the question of why Sorbian is not a canonical null-subject language. Within a licensing/identification system it would be necessary to answer the question of why the identification of thematic null subjects in Sorbian is blocked despite the obviously *rich' verbal morphology. Two (more or less successful) models were then formulated. One was derived from the possibility that the number of syncretisms allowable for successful identifica-

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-197- tion had been exceeded, and the other was based on word order, specifically the verb-final preference found in Sorbian, which is peculiar in Slavic. However, both of these models presented potential difficulties in the handling of certain data, or produced counter-intuitive predictions. In the final analysis it was determined that the licensing/identification system in its restricted form cannot account for the Sorbian data. As

an alternative, I proposed a model in which Sorbian is analyzed as a truly optional pro-drop language in which pro is

licensed and identified, and where the presence of atonic personal pronouns causes the null and overt variants to be in competition. In chapter 5 I discussed the occurrence of overt expie-

tives in Czech and Sorbian. This analysis led to the conclusion that these elements, though different in their syntactic status, are actually not a challenge for the predictions of possible language types made by the kind of identification and licensing system suggested by Jaeggli/Safir (1989) in conjunction with the avoid pronoun principle.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access BIBLIOGRAPHY Avanesov R.I. and V.V. Ivanov. (1982) Istori£eskaja grammatika msskogo jazyka. Moskva: Nauka.

Babby, L. (1989) "Subjectlessnesst external subcategorization, and the Projection Principle". Zbornik za filologijų i lingvistika 32: 7-39. Bauer, J. (1968) "Strukturni typy slovanské vëty a jejich vyvoj ". Československė pfednàëky pro VI. mezinárodní sjezd slavistû. Praha: Akademia.

Benedicto, E. (1994) "Agr, 0-features and V-movement: Identifying pro". C/niversity of Massachusetts occasional papers in linguistics 17: 1-18. Bennis, H. and L. Haegemann. (1984) "On the status of agreement and relative clauses in West-Flemish" . De Geest and Putseys eds. Sentential complementation. Dordrecht: Foris, 33-53. Berneker, E. (1900) Die Wortfolge in den slavischen Sprachen. Berlin: Behr*s Verlag.

Bloomfield, L. (1935/1958) Language. New York:Holt and Co. Borer, H. (1989) "Anaphoric AGR". Jaeggli 0. and K.J. Safir eds. The null subject parameter. Dordrecht: Kluwer, 69-109. Bouchard, D. (1989) On the content of empty categories. Dord- recht: Foris. Cardinaletti, A. (1990) nEs pro and sentential arguments in German". Linguistische Berichte 126: 135-164. Carlton, T. (1990) Introduction to the phonological history of the Slavic languages. Columbus,OH: Slavica. Chomsky, N. (1981) Lectures on government and binding. Dord- recht: Foris. Chomsky, N. (1982) Some concepts and consequences of the theory of government and binding. Cambridge,MA: MIT Press. Chomsky, N. (1986) Barriers. Cambridge,MA: MIT Press. Černyx, P.J. (1954) Istoričeskaja grammatika russkogo jazyka. Moskva : Nauka.

De Bray, R. (1963) Guide to the Slavonic languages. London: Dent & Sons.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Die Sorben in der Statistik*1. Die Sorben in״ (Elle, L. (1991 Deutschland. Bautzen: Macica Serbska, 21-25. Elle, L. (1992) "Zur Sprachsituation der Lausitzer Sorben". Slavia Occidentalis 48/49: 67-83. Emonds, J. (1976) A transformational approach to syntax. New York: Academic Press. Faska, H. (1964) ,*Je personalny pronomen pii formje finitneho słowjesa njetrëbny?" Serbska šulą 64: 572-576. Faßke, H. (1979) "Semantika a distribucija personalnych pronomenow w hornjoserbskej spisownej rēči". Lëtopis instituta za serbski ludospyt 26, 1-23. Faßke, H. (1981) Grammatik der obersorbischen Schriftsprache der Gegenwart. Morphologie. (in conjunction with F. Michałk) Bautzen: VEB Domowina-Verlag. Faßke, H. and F. Michałk (1989) "Dialektny tekst z Mješic". Lëtopis instituta za serbski ludospyt 36: 38-52. Feldstein, R. (1987) ”Czyta vs. czytaj and the determination of Polish conjugational desinences" International journal of Slavic linguistics and poetics XXXV-XXXVI: 65-76. Franks, S. (1990) "On the status of null expletives". Lingua 81: 1-24.

Franks, S. (1994) "Null subject characteristics of Slavic languages". Paper presented at the annual meeting of the American Association of Teachers of Slavic and Eastern European Languages. Franks, S. (1995) "Null subject phenomena". Parameters of Slavic morphosyntax. Cambridge: Oxford University Press. Franks, S. and G. Greenberg. (1993) "The functional structure of Slavic clauses". Forrna! approaches to Slavic linguistics: The Ann Arbor meeting. Ann Arbor: Slavica 77-108. Gribble, C. (1973) Medieval Slavic texts. Cambridge,MA: Slavica. Haegeman, L. (1994) Introduction to government & binding theory. Cambridge,MA: Basil Blackwell. Haider, H. (1991) "PRO-PLEME". Strukturen und Merkmale syntaktischer Kategorien. Tübingen: Gunter Narr. (Studien zur deutschen Grammatik, 39.)

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access -200- Hammer, L.В. (1985) Prague colloquial Czech. Ph.D. dissertation, Indiana University.

Hirschová, M. (1984) "K jednomu typu vypovëdl bēžnē mluvené .Slovo i Slovesnost 45: 273-275 .״češtiny

Holmberg, A, Nikanne U. (1994) "Expletives and subject positions in Finnish". Proceedings of NELS 24: 73-187. Huang, C.-T. J. (1984) "On the distribution and reference of empty pronouns”. Linguistic inquiry 15: 531-574. Hyams, N. (1986) Language acquisition and the theory of parameters. Dordrecht: Reidel. Innis, J. et al. (1985) Intermediate and advanced Russian reader. Ann Arbor: Ardis.

Ivanov, V .V . (1980) Istoričeskaja grammatika russkogo jazyka. Moskva : Prosveščenie. Jaeggli, 0. (1982) Topics in Romance syntax. Dordrecht: Foris. Natural .״Jaeggli, O. (1986) "Arbitrary plural pronouns language and linguistic theory 4: 43-76.1 Morphological uniformity and״ (Jaeggli, O. and N. Hyams. (1987 the setting of the null subject parameter". Proceeding of NELS 18: 238-253. Jaeggli, O. and K. Safir. (1989) Null subject parameter. Dord- recht: Kluwer. -Denkschrif .״Jagić, V. (1900) "Beiträge zur slavischen Syntax ten der Kaiserlichen Akademie der Wissenschaften. XLVI : 1-88. Wien: Carl Gerold's Sohn. Bally, C. ed. Mélanges de .״Signe Zéro״ (Jakobson, R. (1939 Linguistique. Geneve: Georg et Cie. Selected writings ״Jakobson, R. (1948) "Russian conjugation II: 119-129. Mouton: The Hague. Janaš, P. (1984) Niedersorbische Grammatik. Bautzen: VEB Domowi na-Ve r1ag. Jenč, R. (1959) "Mëstno finitnych formow pomocneho słowjesa a participa wuznamoweho słowjesa w hornjoserbskej sadźe". Lëtopis instituta za serbski ludospyt 6: 3-47. K prašenju normy w hornjoserbskej spisownej״ (Jenč, R. (1964 rēči". Serbska Šulą 17, 1964.4: 247-254.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Jenč, R. (1965) "К praâenju normy w hornjoserbskej spisownej .Serbska šulą 17, 1965.5: 309-317 .״rēči Jentsch, H., Michałk S. 1971. Sorbische Dialekttexte. IX Sàuritz, Ostro, Jauer, Kreis Kamenz. Bautzen: VEB Domowina-Verlag.

Jonas, D. and J.D. Bobaljik (1993) "Specs for subjects: The role of TP in Icelandic". MIT working papers in linguistics 18: 59-98.

Jordan, F.P. (1841) Grammatik der sorbisch-wendischen Sprache. Prag: Friedrich Ehrlich. King, T. (1993) Configuring topic and focus in Russian. Ph.D. dissertation, Stanford University.

Koch, J. (1991) Ha lećała je módra wróna. Bautzen: VEB Domowina-Verlag. Koenitz, В. (1988) "Zur Charakteristik der Satzformen mit "expletivem" Personalpronomen im Tschechischen und Deutschen". Linguistische Arbeitsberichte 66: 8-14, Karl- Marx-Universität Leipzig. Koktová, E. (1992) "Anaphoric expressions across clause boundary in pro-drop languages". Prague bulletin of mathematical linguistics 58: 39-63.

Kosta, P. (1990) "Zur formalen Lizensierung und inhaltlichen Determination leerer Subjekte im Russischen". Slavisti- sehe Beiträge 260: 117-165. München: Verlag Otto Sagner. Kral, G. (1895) Gra/nmatiJc der Wendischen Sprache in der Ober- lausitz. Bautzen: Druck und Verlag von Schmaler. Kroch, A. (1989) "Function and grammar in the history of English: Periphrastic ÜQ." - Fasold, R. and D. Schiffrin (eds). Language chćmge and variation. Amsterdam: Benjamins. Kuen, H. (1957) "Die Gewohnheit der mehrfachen Bezeichnung des Subjekts in der Romania und die Gründe ihres Aufkommens". Syntactica und Stilistica. (Festschrift für Gamillscheg) . Tübingen: Max Niemeyer.

Lamprecht, A. et al. (1986) Historická mluvnice češt iny. Praha: Státní pedagogické nakladatelství. Leinonen, M. (1983) "Generic zero subjects in Finnish and Russian". Scando-Slavica 29: 143-161.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access Liebsch, G. (1884) Syntax der wendischen Sprache in der Oberlausitz. 1973 reprint. Bautzen: VEB Domowina-Verlag. Lindseth, M. (1994) "Expletive and thematic subjects in West Slavic". Studies in linguistic sciences 24.2: 333-346. Lindseth, M. and S. Franks (1996). "Licensing and identification of null subjects in Slavic". In Formal approaches to Slavic linguistics : The College Park meeting, 199-225. Lindseth, M., Ludwig, J. (1992) "The overt, non-referential pronoun in Czech, Lusatian and German". Paper presented at the annual meeting of the American Association of Teachers of Slavic and Eastern European Languages. Lorenz, K. (ed.) (1981) Serbska čitanJca. Leipzig: Reclam. Lötzsch, R. (1968) "Einige Auswirkungen des Purismus auf die grammatische Normierung slawischer Schriftsprachen". SoraJbistisJce pfinoëki к VI. mjezynarodnemu kongresej slawistow w Praze 1968. Spisy instituta za serbski ludospyt 26. Bautzen: VEB Domowina-Verlag. Meier, G.F. (1961) Das Zero-Problem in der Linguistik. Berlin: Akademie-Verlag. Michałk, F. (1956-57) "Slowoslëd w Serbščinje". Lëtopis instituta za serbski ludospyt 17: 1-29. Michałk, F. (1962a) "Wliw nëmCiny na poioženje finitneho słowesa w serbskej sadźe" Otázky slovanské syntaxe. Praha: Statní Pedagogické Nakldatelství. Michałk, F. (1962b) "Der Einfluß des Deutschen auf die Stellung des Verbum finitum im sorbischen Satz". Zeitschrift für Slavistik 62, 232-261. Michałk, F. (1965a) "Wliw serbščiny na syntaksu пётбіпу we Lužicy". Serbska šulą 18: 711-718. Michałk, F. (1965b) "Enkliza w serbŠčinje". Serbska šulą 18: 124-128. Michałk, F. (1970) "K prašenjam slowoslëda w serbskich dialek- tach". Lëtopis instituta za serbski ludospyt 17: 1-29. Michałk, F. (1972) "Latinizmy a germanizmy w rēči J. H. Swët- lika". Lëtopis instituta za serbski ludospyt 19: 52-106. Michałk, S. (1977) "Typische Erscheinungen des sorbisch- deutschen Sprachkontakts". Methodische Beiträge zum Sprachunterricht 1977: 75-105.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-203- Michalk, S. and H. Protze. (1974) Studien zur sprachlichen Interferenz. Deutsch-sorbische Dialekttexte aus Radibor, Kreis Bautzen. Bautzen: VEB Domowina-Verlag. Miklosich, F. (1883) Subjectlose Sätze. Wien: Wilhelm Braumüller. -Jaeggli, О. and Silva .״?How Pro is it״ (Montalbelli, M. (1986 Corvalan, C. eds. Studies in Romance linguistics. Dord- recht: Foris. Müller, G. (1989) "Eine Geschichte ohne Subjekt. Zur Entwick- lung der pro-Theorie". Linguistische Berichte 119: 3-52. Nakhimovsky A.D. and R.L. Leed (1987) Advanced Russian. Columbus,OH: Slavica Publishers. Nowak, H. (1988) Powëdamy dolnoserbski. Gutes Niedersorbisch. Bautzen: VEB Domowina-Verlag.

Oreânik, Janez. (1986) "The obligatorium of unemphatic pronoun subjects in the Germanic languages". Filologija 14: 261- 269. Oschlies, W. (1991) Die Sorben - Slawisches Volk im Osten Deutschlands. Bonn: Friedrich-Ebert-Stiftung. Perlmutter, D.M. (1971) Deep and surface structure constraints in syntax. New York: Holt, Rinehart and Winston. Pëtrancec, M. (1978) '1Bjezsubjektowe werbalne sady w ruščinje a j ich pfeložk do serbščiny". Diplomarbeit, Leipzig. Pollock, J.Y. (1989) "Verb movement, universal grammar, and the structure of IP". Linguistic inquiry 20: 365-424. On the discourse functions of syntactic״ (Prince, E. F. (1993 form in Yiddish: Expletive ES and subject-postposing". Goldberg D. ed. The field of Yiddish. Evanston: Northwestern University Press. Rizzi, L. (1982) Jssues in Italian syntax. Dordrecht: Foris. (Chapter IV, "Negation, wh-movement and the null subject parameter") Rizzi, L. (1986) "Null objects in Italian and the theory of pro". Linguistic inquiry 17: 501-557. Roberge, Y. (1990) The syntactic recoverability of null arguments. Kingston: McGi11-Queen's University Press. Roberts, I. (1993) Verbs and diachronic syntax. Dordrecht: Kluwer.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-204- Explaining the syntactic consequences״ (Rohrbacher, В. (1995 of "rich" agreement morphology: On the licensing of V-to- AgrS raising and pro'*. Proceedings of WCCFL 13: 350-364. Safir, K. (1985) Syntactic chains. Cambridge University Press. Schalich, G. (1970) Expressive sprachliche Mittel im modernen tschechischen Lustspiel. München: Fink Verlag. Schneider, F. (1958) Grammatik der wendischen Sprache katholi- sehen Dialect*s. Bautzen: Hiecke. Schuster-Šewc, H. (1967) Sorbische Sprachdenkmäler: 16.-18. Jahrhundert. Bautzen: VEB Domowina-Verlag. Schuster-Šewc, H. (1974) "Sätze mit fiktiven Subjekt vom Typ os. ,Wono so deščuje /ns. to se pada1 und ihre Stellung :Zeitschrift für Slawistik 19 .״in der slawischen Syntax 340-352.

Schuster-Šewc, H. (1976) Gramatika hornjoserbskeje rëöe. Syntaksa. Bautzen: VEB Domowina-Verlag. Seiler, A. (1830) Kurzgefaßte Greunmatik der sorbischen Spra- che. (Fotomechanischer Neudruck Bautzen: VEB Domowina- Verlag, 1973). Short, D. (1994) "Czech". Comrie B. & G. Corbett eds. The Slavonic languages. London: Routledge, 455-532. Siebenschein, H. (1988) Deu tsch - Tschechisches/Tschechisch Deutsches Wörterbuch. Praha: Stâtnl Pedagogicke Nakladatelstvi. Smoler, J.A. (1841) Mały Serb. Wendisch-deutsche Gespräche. Bautzen: Wellersche Buchhandlung. Speas, M. (1994) "Null arguments in a theory of economy of projection". University of Massachusetts occasional papers in linguistics 17: 179-208. Sperling, A. (1990) "Arbitrary pronominais in Slavic". Unpublished manuscript, Indiana University. Spieß, G. (1995) "Die sprachliche Situation bei den Sorben in der Niederlausitz". Lëtopis instituta za serbski ludospyt 42: 59-62, Sonderheft. Sportiche, D. (1988) "A Theory of floating quantifiers and its corollaries for constituent structure". Linguistic inquiry 19: 425-449.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

־-205 Sprouse, R. and В. Vance (1995) "An Explanation for the decline of null pronouns in certain Germanic and Romance languages". To appear in Language creation and change: Creolization, diachrony, and development. MIT Press. Śołta, В. (1921) "Prëni hornjoserbski pfeložk biblije". Časopis Maćicy Serbskeje, 3-34. Śołta, J. et al. (1984) Nowy biogrāfiski słownik. Bautzen: VEB Domowina-Verlag. Stone, G. (1994) "Sorbian". Comrie B. & G. Corbett eds. The Slavonic Languages. London: Routledge, 593-685. Taraldsen, K.T. (1978) "On the nominative island condition, vacuous application and the that-trace filter". Mimeo- graphed ms., MIT, Cambridge; distributed by the Indiana University Linguistics Club, Bloomington, Indiana, 1980. Tornaseli!, A. (1982) "Das unpersönliche "es" - Eine Analyse im Rahmen der generativen Grammatik" . Linguistische Berichte 102: 171-190. Townsend, C. (1990) A description of spoken Prague Czech. Columbus, OH: Slavica. Trávníőek, F. (1951) Mluvnice spisovné češtiny II. Praha: Slovanské Nakladatelství. Trávnííek, F. (1962) Historická mluvnice öeská III. Praha: Státní Pedagogické Nakladatelství. Tschernych, P.J. (1957) Historische Grammatik der russischen Sprache. Halle: Niemeyer. Waurick, I. (1968) "Der Inhaltssatz im Niedersorbischen". Zur grammatischen und lexikalischen Struktur der slawischen Gegenwartssprachen. Halle: Max Niemeyer, 113-130. Williams, E. (1987) "Implicit arguments, the binding theory, and control". Natural language and linguistic theory 5: 151-180. Woolhiser, C. (1994) "Expletive pronouns in Slavic: Subject or topic?" Indiana University manuscript. Zubatÿ, J. (1917) "Die 1 man'-Sätze". Zeitschrift für verglei- chende Sprachforschung XL: 478-520.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-206- Bibles and Pericopes: Upper Sorbian: (in chronological order) S.Matthaeus und S.Marcus t translated from German by M. Frencel, 1670. Swjate Sćenja, Lekcijony ha Epistle na te Njedźele, ha swjate Dny t oho cyioho Lēta, translated from Latin by J. H. Swëtlik, 1690 (pericopes). Biblija, to je cyłe Swjate Pismo Stareho a Noweho Zakona, translated from German by several Lutheran ministers, 1728. Swjate sćenja, lekciony a epistle ... 1750 (pericopes) Biblia, to je cyłe Swjate Pismo Stareho a Noweho Zakona, translated from German by Kühn 1823. Swjate sćenja, lekciony a epistle ... 1848 (pericopes). Zowy Zakoń našeho knjeza Jézusa Chrystusa, translated by Jakub Buk, 1862. Nowy zakón, do homjosebščiny po rjedźe Vulgaty pfeiožištaj, translated form Latin by J. Lusčanski a M. Hórnik, Budyšin 1896. Perikopy. Epistle, lekcije a sćenja . . . 1909 (pericopes) . Sćenje po swjatym Mateju, originally translated by M. Frencel and newly translated by the Bible commission of Lutheran Sorbian ministers, 1960. Swjate Pismo Noweho Zakonja, 1966.

Lower Sorbian: Das niedersorbische Testament des Miklawuś Jakubica, 1548. (pubi, by H. Schuster-Sewc, Akademie-Verlag : Berlin, 1967.) Nowy Testament, translated by G. Fabricius 1709.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access 00052018

-207- Czech :

Bibif Svatá aneb všecka sva tá pisma Starébo i Nového Zákona Podle posledního vydání Kralického z roku 1613 ("Kralická Bible") Biblí Öeská 1677 ("Wenceslaus Bible").

Biblí Öeská čili Pismo Svaté Starého i Nového Zákona, 1889.

Latin :

Biblia Sacra, Vulgatae editionis, 1914 ("Latin Vulgate").

German :

Die Bibel oder die ganze Heilige Schrift, translated by M Luther, 1522.

Martina Lindseth - 9783954790661 Downloaded from PubFactory at 01/10/2019 02:50:54AM via free access